Lexical Approaches to Argument Structure - hu-berlin.destefan/PS/arg-st.pdf · Lexical Approaches...

Lexical Approaches to Argument Structure

Stefan Müller

Freie Universität Berlin

Stephen Wechsler

University of Texas at Austin

October 19, 2012

CONTENTS 1

Abstract

We conclude that argument structure properties should be representedtogether with lexical items.

Contents

1 Introduction 2

2 Lexicalist approaches 4

2.1 Predicate argument structure . . . . . . . . . . . . . . . . . . . . 42.2 Views on Lexical Rules . . . . . . . . . . . . . . . . . . . . . . . 6

3 Non-lexical approaches 8

3.1 Preliminary theory comparison . . . . . . . . . . . . . . . . . . . 83.2 Non-lexical approaches that refer to phrase structure . . . . . . . 9

4 The pendulum of lexical and phrasal approaches 11

4.1 Some historical notes . . . . . . . . . . . . . . . . . . . . . . . . 114.2 Morphological Derivation . . . . . . . . . . . . . . . . . . . . . . 124.3 Partial Frontings . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

5 Valence versus plugging 17

5.1 Usage-based theories . . . . . . . . . . . . . . . . . . . . . . . . 175.2 Coercion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205.3 Valence and Coordination . . . . . . . . . . . . . . . . . . . . . . 225.4 Valence and Derivational Morphology . . . . . . . . . . . . . . . 285.5 Simplicity and polysemy . . . . . . . . . . . . . . . . . . . . . . 29

6 Radical Underspecification: The End of Argument Structure? 31

6.1 Little v and idiom asymmetries . . . . . . . . . . . . . . . . . . . 336.2 Deverbal nominals . . . . . . . . . . . . . . . . . . . . . . . . . 356.3 Idiosyncratic Syntactic Selections . . . . . . . . . . . . . . . . . 386.4 Expletives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

†We thank Bill Croft, Charles Fillmore, Adele Goldberg, Paul Kay, Jean-Pierre Koenig, andIvan Sag for extended discussions via email and in person and [your name here?] for commentson an earlier version of this paper.

We thank Patience Epps for forwarding our question regarding causatives and passivizationto the LingTyp mailing list and Peter Arkadiev, Anna Bugaeva, Mark Donohue, Diana Forker,Alexander Letuchiy, Yury Lander, Geda Paulsen, Barbara Stiebels, Kashi Wali, and BjoernWiemer for examples and pointers to literature. We are grateful to all replies but decided to useonly some of the data.

Draft of October 19, 2012, 21:14. Comments Welcome!

mailto:[email protected]

2 1 Introduction

6.5 Is there an alternative? . . . . . . . . . . . . . . . . . . . . . . . 436.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

7 Relations between constructions 44

7.1 Inheritance hierarchies for constructions . . . . . . . . . . . . . . 457.2 Mappings between Different Levels of Representations . . . . . . 467.3 Is there an alternative to lexical rules? . . . . . . . . . . . . . . . 49

8 Arguments from Language Acquisition 49

9 Lexical Items, Licensed Trees, Classes of Trees 52

10 Arguments from Psycholinguistics 54

11 Arguments from Statistical Distribution of Material 56

11.1 Unsupervised Data-Oriented Parsing . . . . . . . . . . . . . . . . 5611.2 Collostructions . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

12 Arguments from Neurolinguistics 60

13 Conclusion 64

14 Appendix: subject idioms 65

1 Introduction

Central to the mastery of a language is knowledge of the predicate-argument re-lations: an English speaker interpreting the sentence The rabbit nibbled a carrotknows that a nominal object following the verb nibble represents the food or othersolid substance that is consumed, while a subject preceding it fills the role ofthe consumer of that substance. But the exact nature of that knowledge and howthat information is represented within the grammar, remain matters of controversywithin linguistics. Simplifying the current debate, one can distinguish lexical ver-sus phrasal approaches.1 In this paper we argue for a certain class of lexicalapproaches.

In lexical (or lexicalist) approaches, words are phonological forms paired withvalence structures (also called predicate argument structures). For a certain headword (in our example above, the verb nibble) and certain argument phrases, the

1 The phrasal approaches are usually called constructional, but we use that label cautiouslysince it is also used for approaches that are explicitly lexical. See for instance Kay (2005); Sag(To appear).



3

word’s predicate argument structure specifies the meaning of the combination as afunction of the meanings of the parts. Lexical rules grammatically encode the sys-tematic relations between cognate forms and diathesis alternations. The syntacticcombinatorial rules are usually assumed to be very general and few in number.

In contrast, phrasal (or constructional; but see footnote 1) approaches eschewthe use of lexical rules. Instead, different morphological cognates and diathesisalternants are captured by plugging a single word (or root) into different construc-tions. The construction carries a meaning that combines with the word’s meaning.In some versions the constructions are phrasal structures, while in others, they arenon-phrasal grammatical constructs called argument structure constructions thatresemble the lexicalist’s predicate argument structure, minus the specific verb orother predicator.

The lexical and phrasal approaches differ. The lexicalist’s predicate argumentstructure is an autonomous, reified grammatical entity, while the constructionalapproach seeks to avoid such entities. A phrasal construction or argument struc-ture construction is more like a mechanism for the filling of argument roles dur-ing the generation or interpretation of a sentence. It is ‘grounded’ in actual sen-tences. Also, as noted above, the construction carries a meaning, and so someof the phrasal approaches would replace standard phrase structure rules or syn-tactic valence frames with meaningful constructions. For both of these reasons,constructional approaches are often affiliated with usage-based theories of humanlanguage that deny the existence, or downplay their importance, of ‘meaningless’algebraic syntactic rules such as phrase structure rules defined purely on syntacticcategories like V and NP. On the usage-based view, the progressive generaliza-tion over input patterns that explains language acquisition and use is incapable ofabstraction to the point of removing communicative content entirely (Tomasello,2003). Thus the resolution of the lexical-constructional debate has potentiallybroad theoretical consequences.

Sections 2 and 3 lay out these approaches in more detail and present prelimi-nary arguments in favor of a particular lexical approach. Following that initial out-line of the approaches, Section 4 provides a brief historical overview of the devel-opments in theoretical linguistics of the last century, which shows that the devel-opment progressed in waves oscillating between phrasal and lexical approaches.We discuss the reasons for changes and thereby point to problems that still ex-ist in current approaches, or have been reintroduced into them. Then we revisitsome classic arguments for the lexical approach, present some new arguments, andanswer challenges involving acquisition (Tomasello, 2003; Goldberg, Casenhiserand Sethuraman, 2004), psycholinguistics (Goldberg, 1995, 2006), neurolinguis-tics (Pulvermüller et al., To appear), statistical distribution (Stefanowitsch andGries, 2009; Bod, 2009a,b), and coercion. We conclude in favor of the lexicalapproach.



4 2 Lexicalist approaches

2 Lexicalist approaches

2.1 Predicate argument structure

On lexical approaches, a word’s predicate argument structure, or valence struc-ture, indicates the number and type of arguments, and specifies the meaning ofthe combination of the word and its argument phrases as a function of the mean-ings of the parts. The following entry for the word nibble indicates that when itappears together with certain arguments, the combination has a certain semanticCONTENT:

(1) A predicate argument structure:

PHON 〈 nibble 〉

ARG-ST⟨

NP x , NP y

⟩

CONTENT nibble(agent: x, patient: y)

The rules of syntax specify the positions for ARG-ST list items, thus interactingwith this structure to license a grammatical clause or other phrasal constructionwith the right meaning. Note that the predicate argument structure is slightlyabstract: it does not directly encode the phrase structure or precedence relationsbetween this verb and its arguments. This abstraction captures the commonalityacross different syntactic expressions of the arguments of a given root.

(2) a. The rabbits were nibbling the carrots.

b. The rabbits were nibbling at/on the carrots.

c. The rabbits were nibbling.

d. The carrots were being nibbled (by the rabbits).

e. a large, partly nibbled, orange carrot

f. the quiet, nibbling, old rabbits

g. the rabbit’s nibbling of the carrots

h. The rabbit gave the carrot a nibble.

i. The rabbit wants a nibble (on the carrot).

j. The rabbit nibbled the carrot smooth.

Verbs exhibit variable polyadicity, i. e. direct-oblique and other diathesis alterna-tions (2a,b), argument optionality (2c), and morpholexical operations like passive(2d), as well as antipassive, causative, and applicative in other languages. Theyhave cognates in other parts of speech such as adjectives (2e,f) and nouns (2g,h,i).Verbs have been argued to form complex predicates with resultative secondarypredicates (2j), and with serial verbs in other languages.



2.1 Predicate argument structure 5

On the view pursued here, the same root lexical entry nibble, with the samemeaning, appears in all of these contexts. The effects of lexical rules togetherwith the rules of syntactic mapping dictate the proper argument expression ineach context. For example, if we call the first two arguments in an ARG-ST listlist (such as the one in (1) above) Arg1 and Arg2, respectively, then in an activetransitive sentence Arg1 is the subject and Arg2 the object; in the passive, Arg2 isthe subject. When adjectives are derived from verbal participles, whether active(a nibbling rabbit) or passive (a nibbled carrot), the rule is that whichever rolewould have been expressed as the subject of the verb is assigned by the participialadjective to the noun that it modifies.

Summarizing so far, a predicate argument structure specifies the relation be-tween a head word, its arguments, and the meaning that results when they arecombined. Rules of syntactic mapping specify the way the arguments are realized(or suppressed) in the syntactic environment of the word. Lexical rules specifysystematic relations between predicate argument structures, within the grammarof a language. The main issue we address in this paper is whether such lexicalrules, as described in more detail in the next section, exist within the minds ofspeakers. We argue that they do.

In addition to predicate argument structures, we assume that grammars includemeaningful phrasal constructions. That is, we agree with (Goldberg, 1995; Toma-sello, 2003; Goldberg and Jackendoff, 2004; Jackendoff, 2011) that grammarsshould contain a phrasal component for certain constructions, such as the N-P-Nconstruction of the kind in (3) discussed by Jackendoff (2008) and the verblessdirectives in (4) mentioned by Jackendoff and Pinker (2005, p. 220) and discussedin detail by Jacobs (2008).2

(3) student after student[NP/advP N-P-N]

(4) a. Off with his head!

b. Into the trunk with you!

In addition to cases like (3) and (4), the analysis of some idioms seems to callfor phrasal lexical items, that is, phrases in which more than one word is fixed(Abeillé and Schabes, 1989; Richter and Sailer, 2009). Other classes of idioms canbe handled by analyses in which words select particular lexemes in their valencefeatures (Sag, 2007). Some combination of these two is often posited, in order tocapture the full range of idiom types, from fixed phrases to syntactically analyz-able idioms. See Sailer, 2000; Soehn and Sailer, 2008 for lexical approaches toidioms.

2See G. Müller (2011) for a lexical account of Jacobs’ data and Müller (2010a, Sec-tion 11.11.9.1) for discussion.



6 2 Lexicalist approaches

While we think grammars include meaningful phrasal constructions, we do notthink lexical rules can or should be eliminated by representing argument structurephrasally.

2.2 Views on Lexical Rules

There are various understandings of what lexical rules are. The first dimensionalong which lexical rules could be classified was discussed by Jackendoff (1975).Jackendoff distinguishes between two conceptions: (i) lexical rules that relate twostored lexical entries and thereby capture redundancies in the lexicon; and (ii)lexical rules that license new lexical items. On the latter view, lexical rules canapply to stored lexical items (which are called lexical entries here) or to lexicalitems that are licensed by a lexical rule or a chain of lexical rules.

The second dimension was discussed in the more formal literature on lexi-cal rules in the 90s (Copestake, 1992; Riehemann, 1993, 1998; Calcagno, 1995;Briscoe and Copestake, 1999; Meurers, 2001), but as it turns out there seem to benot just formal but also empirical differences between the approaches (Goldberg,To appear). Calcagno (1995) and Calcagno and Pollard (1995) argued for a viewon lexical rules that was called the meta-level approach by Meurers (2001). Thisapproach can be sketched as in (5):

(5) L1 → L2

Here L1 and L2 are descriptions of lexical objects. The rule states that if the lan-guage contains a lexical object satisfying L1 then it contains another lexical objectsatisfying L2. Thus the rule is not itself a description but a ‘meta-description.’An alternative is the description level approach suggested by Copestake, Briscoe,Riehemann, and Meurers. There are two variants of this approach. Both are shownin (6):

(6) a.

OUT L2

IN L1

lr-type

b.

L2

DTR L1

lr-type

What is shown in (6) is two typed feature descriptions. In both versions L1 is thevalue of a feature. In (6a) the output of the lexical rule is the value of a feature,while it is not in (6b). The lexical rule in (6b) licences an object that can be usedin further analyses since it basically has the properties of L2.3

3For approaches that assume a representation like (6a) additional machinery is needed. Meur-



2.2 Views on Lexical Rules 7

In what follows, we mean (6b) when we speak of lexical rules. As suggestedby the name of the feature DTR, which stands for DAUGHTER, this type of lexicalrule is equivalent to a unary branching tree. In Figure 1 a lexical rule licenses thesubtree where the daughter node is the verb BREAK and the mother has the 3rdperson singular inflected form (breaks) and agreement features. In Figure 2 on thenext page the passive lexical rule licenses a structure with the same verb BREAK

is the sole daughter and the mother has the passive participle form and a valencefeature in which the role of the object of the active is assigned to the subject ofpassive. Figure 3 on page 9 illustrates the participle-to-adjective conversion ruleas applied to a passive verbal participle. The effect of this rule is that the semanticrole that the participle would assign to its subject is assigned by the adjective tothe noun it modifies.

S

NP VP

V NP

Kim breaks toys

BREAK

Figure 1: Example of lexical rule for present tense verb.

ers suggested the following word principle to be able to state that lexical items that are licensed bylexical rules can be used as words.

(i) word ⇒ (L1 ∧ STORE 〈〉) ∨ . . . ∨ (L2 ∧ STORE 〈〉 . . . ) ∨ 1

[

STORE

[

OUT 1

lex-rule

]

]

This word principle disjunctively lists all lexical entries and furthermore says that linguistic objectsthat correspond to the output of a lexical rule are words. The 1 indicates structure sharing, that is,idetnity of values. So the linguistic object following the first occurrence of 1 in (6) is identical tothe OUT value of lexical rules like the one in (5a). In this sense the view of lexical rules in (5a)paired with a word principle like (6) is rather similar to the view in (5b). The approaches in (5b)and (5a)+(6) are called template-based approaches by Goldberg (To appear).

Sag (To appear) suggests an approach along the lines in (5a) together with a meta principlethat states that the OUT value is a well-formed linguistic object licensed by the theory that canbe used in other constructions. This approach is similar to the meta level approach in that thelicensed objects do not contain the object from which they are derived. So (4) and Sag’s proposalcorrespond to what Goldberg calls lexical rules.



8 3 Non-lexical approaches

S

NP VP

V VP

Toys were broken

BREAK

Figure 2: Example of passive lexical rule.

3 Non-lexical approaches

Instead of using lexical rules, non-lexical approaches capture morphological cog-nates and diathesis alternants for a single word (or root) by plugging the word intodifferent phrasal constructions. The construction carries a meaning that combineswith the word’s meaning. Phrasal constructions such as the Intransitive, Transi-tive, and Ditransitive constructions replace the phrase structure rules or valenceframes of other syntactic theories; others include the Caused Motion and Resul-tative constructions. The ditransitive construction means ‘X caused Y to receiveZ’ and can combine with either a 3-argument verb like fax (Pat faxed Bill the let-ter) or a 2-argument verb like bake (Pat baked Bill a cake). In the latter case theconstruction licenses the recipient argument.

3.1 Preliminary theory comparison

As noted in the introduction, there is a subtle but real theoretical difference be-tween the lexical and non-lexical constructional approaches. The lexicalist’s pred-icate argument structure is an autonomous, reified grammatical entity that there-fore need not immediately combine with its specified arguments, but can alterna-tively meet other fates: it can serve as the input to a lexical rule; it can combinefirst with a modifier in an adjunction structure or with a similar head in a coordi-nate structure; instead of being realized locally, one or more of its arguments canbe effectively transferred to another head’s valence feature (raising or argumenttransfer); or arguments can be saved for expression in some other syntactic posi-tion (partial fronting). (These phenomena are discussed below.) But on the non-lexical constructional approach, the argument structure construction is more like amechanism for the filling of argument roles during the generation or interpretation



3.2 Non-lexical approaches that refer to phrase structure 9

DP

D NP

AP NP

the brokenA toys

brokenV

BREAK

Figure 3: Example of passivization and adjectivalization lexical rules.

of a sentence. The combination of the verb and argument structure constructiondoes not enjoy an autonomous status in the grammar; it is not abstracted awayfrom its syntactic context, but rather is ‘embodied’ in the full structures found inutterances. Hence it cannot serve as input to other grammatical rules.

On the phrasal approach, a single clause can involve many constructions, in-volving not only basic argument realization but also extraction, raising, and so on.Interactions between these syntactic processes are captured by organizing con-structions into inheritance hierarchies, from which a given sentence can inheritmultiple constructions. These hierarchies have never been described precisely,but have been shown to be problematical when applied to phrasal constructions(Müller, 2010b). Further problems for their application to both phrasal and non-lexical argument structure constructions are discussed in Section 7.1 below.

3.2 Non-lexical approaches that refer to phrase structure

There are two major variants of phrasal approaches. In some versions the con-structions are non-phrasal grammatical constructs called argument structure con-structions (ASCs). An ASC contains roughly the same information as a lexical-ist’s predicate argument structure but without a specific verb or other predicator.Goldberg’s (1995) ASCs contain grammatical relation names like SUBJ, OBJ,and OBL. Hence her ASC closely resembles an LFG functional structure, onlywithout the verb specified. The verb is stored with some of its roles specified asprofiled, which means they are destined for realization as direct grammatical rela-tions (SUBJ or OBJ). Goldberg assumes that her phrasal constructions just specifygrammatical functions that have to be realized together with a certain head. That



10 3 Non-lexical approaches

is, such constructions can be underspecified with regard to linear order. The onlyrequirement is that the parts of the construction have to be realized somewherein a structure (Goldberg, 1995; Goldberg, 2006). How this comes about is notworked out in detail.

Other authors assume phrase structure-like objects, that is, a certain config-uration with part of speech and structural information is paired with a certainmeaning (Alsina, 1996; Goldberg and Jackendoff, 2004; Bergen and Chang, 2005;Culicover and Jackendoff, 2005; Asudeh, Dalrymple and Toivonen, 2008; Jack-endoff, 2011). We illustrate the empirical consequences of the two variants withthe example of verb-particle constructions in Dutch and German. Booij (2002,Section 2; To appear) and Blom (2005), working in the frameworks of Construc-tion Grammar and LFG, respectively, assume that particle verbs are licensed byphrasal constructions (pieces of phrase structure) in which the first slot is occupiedby the particle.

(7) [ X [ ]V ]V’ where X = P, Adv, A, or N

Examples for specific Dutch constructions are:

(8) a. [ af [ ]V ]V’

b. [ door [ ]V ]V’

c. [ op [ ]V ]V’

This suggestion comes with the claim that particles cannot be fronted. However,this claim is wrong for languages like Dutch and German. On Dutch see Hoek-sema, 1991, p. 19, on German Müller, 2002a,b, 2003b, 2007c. Some more fun-damental remarks on introspection and corpus data with relation to particle verbscan also be found in Müller, 2007c; Meurers and Müller, 2009. A German exam-ple is given in (9), several pages of attested examples can be found in the citedreferences and some more complex examples will also be discussed in Section 12on page 62.

(9) LosPART

damitthere.with

gehtwent

esit

schonalready

amat.the

15.15

April.4

April‘It already started on April the 15th.’

Particle verbs are mini-idioms. So the conclusion is that idiomatic expressions thatallow for a certain flexibility in order should not be expressed in phrasal config-urations describing adjacent elements. For some idioms, a lexical analysis alongthe lines of Sag, 2007 seems to be required.5

4taz, 01.03.2002, p. 8, see also Müller, 2005b, p. 313.5Note also that the German example is best described as a clause with a complex internally

structured constituent in front of the finite verb and it is doubtful whether linearization-based



11

4 The pendulum of lexical and phrasal approaches

4.1 Some historical notes

Our historical section starts with Categorial Grammar (Ajdukiewicz, 1935; Steed-man, 2000). Categorial Grammar (CG) is the prototype for a lexical model: everyword (every functor) comes with descriptions of its arguments and the rules thatcombine functors with their arguments are very general and few in number. For in-stance an English transitive verb like know is assigned the lexical entry (s\np)/np.This means that knows takes an NP to its right and an NP to its left. The rulesfor combination do not contain any part of speech information. For instance therule that combines a verb like know with its object has the form X/Y * Y = X.Such general combinatory rules have a component for semantic combination (forinstance, functional application or composition).

Another branch of theoretical linguistics assumed phrase structure rules asbase component in a transformational setting (Chomsky, 1957). While the rules ofCG are binary branching and rather abstract, the early phrase structure rules werenot. There were rules for VPs with ditransitive verbs that had three daughters (forexamples see Chomsky, 1965, p. 72, 96, 107). On some analyses phrase structurerules introduced rich semantic features directly into the phrase structure, suchas CAUSE for causation (Chomsky, 1970), an approach greatly expanded in theGenerative Semantics school (Lakoff, 1969).

There were different answers on the question of how to integrate semanticsinto Generative Grammar: Transformational Grammar started out assigning se-mantics on the level of Deep Structure but problems quickly became apparent,which led to modifications of the framework and to interpretation rules that tookinto account Surface Structure as well (see Bach, 1976 for an overview). An alter-native to the prevalent view in Transformational Grammar was proposed by Mon-tague (1973), who assumed that interpretation is combined with the rules of syn-tactic combination. Bach (1976, p. 184) called this the rule-to-rule assumption.Also in the 1970s other non-transformational theories like TAG (Joshi, Levy andTakahashi, 1975), LFG (Bresnan and Kaplan, 1982), and GPSG (Gazdar, Klein,Pullum and Sag, 1985) were developed and some of them came with detailed se-mantic representations. For instance Gazdar (1982) and Gazdar, Klein, Pullumand Sag (1985, Chapter 10) are very explicit about the semantic representationsand the combination rules for GPSG. They allow for rule-specific semantic in-terpretation and in fact propose a quite specific composition rule for passivized

proposals like the ones in Kathol, 1995, p. 244–248 or Wetta, 2011 can capture this. The issue ofparticle verbs will be taken up in Section 12 again, where we discuss evidence for/against phrasalanalyses from neuro science. Resultative and caused motion constructions will be discussed inSection 5.4.



12 4 The pendulum of lexical and phrasal approaches

sentences (p. 219). That is they share the rule-to-rule assumption.While Montague’s proposal was in the spirit of Categorial Grammar and as-

sumed binary branching structures, GPSG was not: The authors of GPSG assumeclassical context free phrase structure rules, for example a VP rule with a verb andtwo objects on the right-hand side. Uszkoreit (1987) assumes (derived) rules forclauses in German that licences a verb together with all of its arguments, and whileno interpretation rules are given in this book, it is clear that this rule is combinedwith a semantic representation in a fully worked out version of the theory.

The GPSG of the 1980s resembled some current versions of ConstructionGrammar in its adoption of what we call a plugging proposal: a verb that issemantically compatible with a certain phrasal construction is plugged into thisconstruction. Valence information is not represented as part of lexical items inGPSG. Instead lexical items had a number assigned to them and could be insertedinto phrasal rules that had the same number. It is only in interaction between rulesand these numbers that lexical items are paired with certain arguments. For in-stance laugh is of category 2 so it can form a VP if used with rule (10a) and love,of category 3, can form a VP with rule (10b).

(10) a. VP → H[2]b. VP → H[3], NP

(The H stands for head, that is, for the verb in (10)). On this model lexical rulesare impossible because the verb has no valence feature to which lexical rules couldapply. Alternations like the passive, for example, were captured entirely withinthe phrase structure component, through meta-rules that expanded the stock ofphrase structure rules.

In the next subsections we look at the problems that this proposal faced in or-der to understand why it was finally given up and replaced by theories that assumea lexical representation of valence information. We will look at two phenomenahere: morphological derivation, and partial frontings.

4.2 Morphological Derivation

The first problem with the GPSG model is that there are morphological processesthat are sensitive to valence (Müller, 2010a, p. 129). For instance -able derivation(and German -bar derivation) is possible with transitive verbs only:

(11) a. lösbarsolveable

(nominative, accusative)

b. vergleichbarcomparable

(nominative, accusative, PP[mit])



4.3 Partial Frontings 13

c. * schlafbarsleepable

(nominative)

d. * helfbarhelpable

(nominative, dative)

The verbs have to have at least a nominative and an accusative argument (11a,b),intransitive verbs like sleep or help do not allow for the -bar derivation.

Moreover, it will not work to say that -bar derivation applies only to verbswith a certain number. For example, lösen (‘to solve’) and vergleichen (‘to com-pare’) have different valence frames. This means that a GPSG rule for -bar deriva-tion would have to mention several numbers that correspond to different valenceframes that allow for -bar derivation. Since the numbers by themselves do notcontain any information about the presence of a direct object, such a formulationof the -bar derivation rule would amount to stipulating a seemingly arbitrary setof numbers, and thereby miss an important generalization. This should be com-pared to models that assume a lexical representation of valence: the -bar suffixcan be specified to attach to verbs whose valence list starts with a nominative andan accusative and hence the generalization is captured easily in such models.

4.3 Partial Frontings

Another reason for needing valence information is to allow for variation in wherein the sentence structure the arguments are discharged. For example, Germanallows for partial frontings like (12):

(12) a. [Erzählen]tell

wirdwill

erhe.NOM

seinerhis

Tochterdaughter.DAT

eina

Märchenfairy.tale.ACC

können.can

‘He will be able to tell his daughter a fairy tale.’

b. [Eina


erzählen]tell

wirdwill

erhe

seinerhis

Tochterdaughter.DAT

können.can

c. [Seinerhis

Tochterdaughter.DAT

eina


erzählen]tell

wirdwill

erhe.NOM

können.can

The non-finite verb erzählen may be realized together with all its complements(12c) or with proper subsets of its complements (12a,b) in the so-called prefieldto the left of the finite verb (subjects can also be fronted with non-finite verbs,but this is rather restricted). The problem for GPSG-like approaches is that thearguments are licensed by a certain phrase structure rule. To be able to analyze(12a) and (12b) one needs phrase structure rules that license the verb without anyargument and with a single argument, respectively. In addition it has to be ensuredthat the arguments that are missing in the prefield are realized in the remainder of




the clause. It is not legitimate to omit obligatory arguments or realize argumentswith other properties like a different case, as the examples in (13) show:

(13) a. Verschlungendevoured

hathas

erhe

esitACC

nicht.not

‘He did not devour it.’

b. * Verschlungendevoured

hathas

erhe

nicht.not

c. * Verschlungendevoured

hathas

erhe

ihmhimDAT

nicht.not

The obvious generalization is that the fronted and unfronted arguments must addup to the total set belonging to the verb.

Nerbonne (1986) and Johnson (1986) suggest GPSG analyses that can dealwith the data. However, they assume a valence representation that uses binaryfeatures like NPacc and NPdat. This makes it possible to represent the fact thatthe accusative object is realized in the prefield in (12b) and may not be realizedin the remainder of the clause (in the so-called middle field). Similarly the dativeobject in (12b) is realized in the middle field and hence may not be realized inthe prefield. As both authors state clearly, this incorporates ideas from CategorialGrammar into GPSG. Theories like HPSG (Pollard and Sag, 1987, 1994) that weredeveloped after GPSG, explicitly borrow from CG and hence take over the CGsolution to the partial fronting problem that was developed by Geach (1970). Seefor instance (Pollard, 1996; Meurers, 2000; Müller, 1996, 2002a; Kathol, 2001)and also (Nerbonne, 1994).

If one does not want to go with the lexical specification of valence frames,there seem to be just two alternatives: remnant-movement analysis as often as-sumed in the transformational literature (G. Müller, 1998) and linearization-basedapproaches that allow for discontinuous constituents (Reape, 1994). In remnant-movement-based approaches it is assumed that the prefield is filled by a VP. Theelements that are not realized in the prefield are moved out of the VP before the(remnant of the) VP is fronted. Such movement-based analyses are usually notassumed in non-transformational frameworks,6 but apart from theoretical consid-erations there are also empirical facts that argue against remnant movement. SeeHaider, 1993, p. 281, De Kuthy, 2002, Chapter 4.2.5, De Kuthy and Meurers,2001, Section 2, and Fanselow, 2002 for details.

The linearization-based variant seems to be what Goldberg (2006, p. 10) has inmind when she writes that argument structure constructions are fused with otherconstructions like a VP construction. To our knowledge the details of such an

6See Hinrichs and Nakazawa, 1994 for a notable exception. This work shows that it would bepossible in principle to assume remnant movement analyses.



4.3 Partial Frontings 15

analysis have not been worked out within the Construction Grammar setting, sowe will discuss explicit proposals in other frameworks: The linearization proposalby Reape (1994) was criticized by Kathol (2001, Section 8.6), who argued on thebasis of agreement, case assignment, and passive for a CG-like analysis of Germanverbal complexes. Reape assumed that a raising verb like scheinen (‘to seem’)embeds a full clause and allows for a discontinuous linearization of the parts ofthis clause. Similarly verbs that allow for the formation of a verbal complex as forinstance the control verb verprechen (‘to promise’) allowed the parts of its verbalargument to be serialized discontinuously. Kathol argued that such an approachfails to capture local agreement relations between the finite verb and the subjectof a clause that is embedded under a raising verb. Consider his examples in (14):

(14) a. Denthe

Mädchengirls.DAT

scheintseem.3SG

/* scheinenseem.3PL

schlechtill

zuto

werden.become

‘The girls seem to be getting ill to the stomach.’

b. Duyou

scheinstseem.2SG

/* scheintseem.3SG

nichtnot

zuto

verstehen.understand

‘You don’t seem to understand anything.’

The problem with a purely linearization-based account is that the inflection of theverb that is the head of the subject does not reflect any agreement, since it is aninfinitive with zu. Instead we have agreement with the finite verb one level up. Theexample in (14a) shows that there does not have to be a subject at all. This is dueto the fact that German allows for subjectless predicates and that raising verbs donot care whether the downstairs predicate selects for a subject or not. An approachthat assumes that du (‘you’) is an argument of scheinen (‘seem’) can account forthe agreement relation locally. Similarly, there are so-called remote passives inGerman. The object of a deeply embedded verb gets assigned nominative (Höhle,1978, p. 175–176):

(15) weilbecause

derthe

Wagencar.NOM

oftoften

zuto

reparierenrepair

versuchttried

wurdewas

‘because many attempts were made to repair the car’

This is explained by an analysis that assumes that zu reparieren versucht behaveslike a complex word with respect to passive and hence the accusative object of zureparieren versucht has to be realized as nominative. See (Kathol, 1994; Pollard,1994; Müller 1999, Chapter 15.3.6; 2002a, Chapter 3.2.5) for argument compo-sition analyses. Kathol’s criticism of linearization-based analyses was relativizedlater by the finding that the integration of data like (16) into the analysis of non-finite complementation requires some non-locality:




(16) a. [Einan

Außenseiteroutsider.NOM

gewonnen]won

hathas

hierhere

nochstill

nie.7

never‘No outsider has yet won here.’

To assign the proper case in the prefield and to account for agreement with the NPin the prefield this information has to be accessible to the finite verb. However, ifgewonnen has the category s\np and ein Außenseiter gewonnen has the categorys, the information about the np is not accessible any longer when ein Außenseitergewonnen is combined with the finite verb. This shows that analyses that arebased on the notion of argument composition alone are not sufficient (Meurers,2000; Meurers and De Kuthy, 2001). This means that more like a basic CategorialGrammar analysis or an analysis under the assumptions of Pollard and Sag (1987)and Pollard and Sag (1994) is needed (Kathol, 2003, p. 27).

That the problem of partial fronting is non-trivial and that simply referring tofusion with other constructions is not sufficient is also indicated by data like theDanish example of partial fronting in (17a) that is due to Holmberg (1999), (17c)– which is taken from Müller and Ørsnes, 2012 – provides a corpus example:

(17) a. ? Læstread

harhas

BjarneBjarne

denit

ikke.not

‘Bjarne did not read it.’

b. * Læstread

harhas

BjarneBjarne

ikkenot

bogenbook.DEF

‘Bjarne did not read the book.’

c. menbut

heltwholly

[udelukke]exclude

kancan

manyou

[det]it

dathen

ikkenot

elleror

hvad8

what‘but you cannot wholly exclude it, can you?’

Danish marginally allows for partial frontings but only if the element that is miss-ing from the fronted VP is a pronoun as in (17a). Pronouns have to be serializedto the left of the negation (18a) and full NPs to the right of it as in (18b).

(18) a. BjarneBjarne

læserreads

denit

ikke.not

‘Bjarne is not reading it.’

b. BjarneBjarne

læserreads

ikkenot

bogen.book.DEF

‘Bjarne is not reading the book.’

The point about the examples in (17a) and (17b) is that the linearization of thematerial with respect to each other and with respect to traditional topological fields

7Haider, 1990, p. 96.8http://hope.pointblog.dk/svaert-at-vide-.html, 26.3.2012.


http://hope.pointblog.dk/svaert-at-vide-.html


17

for Danish (Diderichsen, 1957; Ørsnes, 2009) is correct, but still (17a) is possible,but (17b) is not. Any proposal that tries to account for the difference between(17a) and (17b) has to explain why the VP can be discontinuous if the object isa pronoun but not when the object is a full NP. Müller and Ørsnes (To appear,2012) present a fully worked out approach that uses argument composition, but itis difficult to see how a linearization- or fusion-based approach could account forthe data in a non-stipulative way.

Concluding this section, we can say that phrasal models have been proposedearlier in the last century and that the development runs in waves. There have beenarguments against GPSG-like analyses that are still valid and the problems are notaddressed by current phrasal approaches. On the other hand there is evidence thatpurely lexical approaches in the spirit of basic Categorical Grammar without anycomplex valence representations are not sufficient either.

5 Valence versus plugging

We have seen how the pendulum has swung between lexical and phrasal ap-proaches. Of the theories on offer, the best place for that pendulum to come torest, in our view, is at a theory in which words are equipped with valence infor-mation that is subject to the effects of lexical rules formulated as unary branchingtrees (see Section 2.2 above). The previous section reviewed earlier argumentsfor needing lexical representations of valence. In this section we present moredetailed arguments specifically directed against the claim that lexical valence rep-resentations (i. e. predicate argument structures) can or should be replaced by whatwe call a plugging proposal, that is, a system in which a verb or other predicatoris plugged into a meaningful construction.

As noted in Section 2.1 above, we believe that grammars include meaningfulphrasal constructions. Our purpose is not to argue against their existence, butrather to argue that they cannot replace lexical valence representations. In the nextsection we begin by examining claims to a purported advantage of ConstructionGrammar over lexical rules. Then we turn to positive arguments for lexical rules.

5.1 Usage-based theories

For many practitioners of Construction Grammar, their approach to syntax isdeeply rooted in the ontological strictures of usage-based theories of language(Langacker, 1987; Goldberg, 1995; Croft, 2001; Tomasello, 2003). Usage-basedtheorists oppose the notion of ‘linguistic rules conceived of as algebraic proce-dures for combining symbols that do not themselves contribute to meaning.’ (To-masello, 2003, p. 99) All linguistic entities are symbolic of things in the realm



18 5 Valence versus plugging

of denotations; ‘all have communicative significance because they all derive di-rectly from language use.’ (ibid) Although the formatives of language may berather abstract, they can never be divorced from their functional origin as a tool ofcommunication. The usage-based view of constructions is summed up well in thefollowing quote:

The most important point is that constructions are nothing more orless than patterns of usage, which may therefore become relativelyabstract if these patterns include many different kinds of specific lin-guistic symbols. But never are they empty rules devoid of semanticcontent or communicative function. (Tomasello, 2003, p. 100)

Thus constructions are said to differ from grammatical rules in two ways: theymust carry meaning; and they reflect the actual ‘patterns of usage’ fairly directly.

Consider first the constraint that every element of the grammar must carrymeaning, which we call the semiotic dictum. Do lexical or phrasal theories hewthe most closely to this dictum? Categorial Grammar, the paradigm of a lexicaltheory (recall Section 4), is a strong contender: it consists of meaningful words,with only a few very general combinatorial rules such as X/Y * Y = X. Giventhe rule-to-rule assumption those combinatorial rules specify the meaning of thewhole as a function of the parts; whether that counts as meaningful in itself is notclear.

What does seem clear is that the combinatorial rules of Construction Gram-mar, such as Goldberg’s Correspondence Principle for combining a verb with aconstruction (1995, p. 50), have the same status:

(19) The Correspondence Principle: Each participant that is lexically profiledand expressed must be fused with a profiled argument role of the construc-tion. If a verb has three profiled participant roles, then one of them may befused with a non-profiled argument role of a construction.

Both verbs and constructions are specified for participant roles, some of which areprofiled. Argument profiling for verbs is ‘lexically determined and highly conven-tionalized’ (Goldberg, 1995, p. 46). Profiled argument roles of a construction aremapped to direct grammatical functions, i. e., SUBJ, OBJ, or OBJ2. By (19) thelexically profiled argument roles must be direct, unless there are three of them, inwhich case one may be indirect.9 With respect to the semiotic dictum, the Corre-spondence Principle has the same status as the Categorial Grammar combinatorialrules: a meaningless algebraic rule that specifies the way to combine meaningfulitems.

9We assume that the second sentence of (19) provides for exceptions to the first sentence.



5.1 Usage-based theories 19

Turning now to the lexicalist syntax we favor, some elements abide by thesemiotic dictum while others do not. Lexical valence structures clearly carrymeaning since they are associated with particular verbs. Also, the lexical rulethat adds a benefactive recipient argument to a verb adds meaning. But phrasestructure rules for intransitive and transitive VPs (or the HPSG ID schemata) donot.

Which structures have meaning is an empirical question for us. For exam-ple, Wechsler (1991, p. 111ff; 1995, p. 88ff) proposed a constructional analy-sis of English ditransitives that prefigures that of Goldberg (1995). Inspired byKiparsky’s (1987; 1988) notion of a thematically restricted positional linker, to-gether with cross-Germanic comparison, the idea was that the inner object posi-tion (the phrasal position of the first of two NP objects in a ditransitive) is seman-tically restricted to expressing the role of ‘intended recipient’: hence He carvedher a toy entails that he carved a toy with the intention that she receive it. Verbslike carve were treated as 2-place, verbs like give as 3-place, and the ‘recipi-ent’ semantics is contributed by the structure itself, redundantly in the case ofgive. Wechsler (1991, Chapter 3; 1995, Chapter 3) treats this object function asa ‘covert dative’ in the Germanic languages that lost dative case, including En-glish. On the other hand, no special meaning was assumed for monotransitive orintransitive structures.

By contrast, in Construction Grammar meaning is assumed a priori for allconstructions. But while the ditransitive construction plausibly adds meaning, nomeaning has yet been discovered for either the intransitive or (mono)transitiveconstructions. In short, the constructionist’s evidence for the meaningfulness ofcertain constructions such as the ditransitive does not constitute evidence that allphrasal constructions have meaning.

Now consider the second usage-based dictum, that the elements of the gram-mar directly reflect patterns of usage, which we call the transparency dictum.The Construction Grammar literature often presents their constructions informallyin ways that suggest that they represent surface word order patterns: the transi-tive construction is ‘X VERB Y’ (Tomasello) or ‘Subj V Obj’ (Goldberg, 1995,2006)10; the passive construction is ‘X was VERBed by Y’ (Tomasello, 2003,p. 100) or ‘Subj aux Vpp (PPby)’ (Goldberg, 2006, p. 5). But a theory in whichconstructions consist of surface patterns was considered in detail and rejected by(Müller, 2006, Section 2), and does not accurately reflect Goldberg’s actual the-ory.11 The more detailed discussions present argument structure constructions,

10Goldberg et al. (2004, p. 300) report about a language acquisition experiment that involves anSOV pattern. The SOV order is mentioned explicitly and seen as part of the construction.

11This applies to argument structure constructions only. In some of her papers Goldberg as-sumes that very specific phrase structural configurations are part of the constructions. For instancein her paper on complex predicates in Persian (Goldberg, 2003) she assigns V0 and V categories.




which are more abstract and rather like the lexicalists’ grammatical elements (orperhaps an LFG f-structure): the transitive construction resembles a transitivevalence structure (minus the verb itself); the passive construction resembles thepassive lexical rule.

With respect to fulfilling the desiderata of usage-based theorists, the differencebetween the non-lexical and lexical approaches is somewhat exaggerated, in ouropinion. However, a difference can be seen by considering novel uses such asHe smiled her his answer or even He carved her a toy.12 On the constructionalapproach, the extra argument(s) of the verbs smile and carve arise only in thecourse of constructing the sentence and combining the verbs with the nominalsthat fill those argument positions. There is no sense in which the verbs have threearguments prior to generating the sentence. Nor is there any substructure of thesentence that corresponds (phonologically) to just the verb’s pronunciation, butencodes the existence of the three arguments. Goldberg’s theory, for example,sedulously avoids positing such entities. This is favorable to the transparencydictum: assuming the locution is novel, then the speaker has never been exposed toa VP of the form [carve NP NP]; but she has been exposed to the verb carve (withjust one object) and to the ditransitive construction. But on the lexical approachthere is a unary branching node with the phonology carve and a 3-place valencestructure. In Section 5.3 we provide evidence for the lexical approach, after firstreviewing a lexical account of such novel utterances in Section 5.2. This evidenceundercuts a usage-based theory that entails the opposite conclusion.

5.2 Coercion

As noted already, researchers working with plugging proposals usually take coer-cion as showing the usefulness of phrasal constructions. For instance, Anatol Ste-fanowitsch (Lecture in the lecture series Algorithmen und Muster — Strukturen inder Sprache, 2009) discussed the example in (20):

(20) Das Tor zur Welt Hrnglb öffnete sich ohne Vorwarnung und verschlang[sie] . . . die Welt Hrnglb wird von Magiern erschaffen, die Träume zuRealität formen können, aber nicht in der Lage sind zu träumen. Haltetaus, Freunde. Und ihr da draußen, bitte träumt ihnen ein Tor.13

The crucial part is ihr träumt ihnen ein Tor (‘Dream a gate for them’). In thisfantasy context the word träumen, which is intransitive, if forced into the ditran-sitive construction and therefore gets a certain meaning. This forcing of a verb

See Müller, 2010b, Section 4.9 for the inadequacies of this analysis.12The former examples are from Tomasello, 2003, p. 160.13http://www.elbenwaldforum.de/showflat.php?Cat=&Board=Tolkiens_Werke&

Number=1457418&page=3&view=collapsed&sb=5&o=&fpart=16 . 27.02.2010.


http://www.elbenwaldforum.de/showflat.php?Cat=&Board=Tolkiens_Werke&Number=1457418&page=3&view=collapsed&sb=5&o=&fpart=16

http://www.elbenwaldforum.de/showflat.php?Cat=&Board=Tolkiens_Werke&Number=1457418&page=3&view=collapsed&sb=5&o=&fpart=16


5.2 Coercion 21

corresponds to overwriting properties of the verb properties by the phrasal con-struction.

In cases in which the plugging proposals assume that information is over-written or extended, lexical approaches assume mediating lexical rules. Briscoeand Copestake (1999, Section 4) have worked out a lexical approach in detail.14

They discuss the ditransitive sentences in (21), which either correspond to theprototypical ditransitive construction (21a) or deviate from it in various ways.

(21) a. Mary gave Joe a present.

b. Joe painted Sally a picture.

c. Mary promised Joe a new car.

d. He tipped Bill two pounds.

e. The medicine brought him relief.

f. The music lent the party a festive air.

g. Jo gave Bob a punch.

h. He blew his wife a kiss.

i. She smiled herself an upgrade.

For the non-canonical examples they assume lexical rules that relate transitive andintransitive verbs to ditransitive ones and contribute the respective semantic infor-mation or the respective metaphorical extension. The example in (21i) is rathersimilar to the träumen example discussed above and is also analyzed with a lexicalrule (page 509). Briscoe and Copestake note that this lexical rule is much more re-stricted in productivity than other lexical rules that were suggested by them. Theytake this as motivation for developing a representational format in which lexicalitems (including those that are derived by lexical rules) are associated with prob-abilities, so that differences in productivity of various patterns can be captured.

Looking narrowly at such cases, it is hard to see any rational grounds forchoosing between the phrasal analysis and the lexical rule. But if we broadenour view, the lexical rule approach can be seen to have much wider application.Coercion is a very general pragmatic process, occurring in many contexts whereno construction seems to be responsible (Nunberg, 1995). Nunberg cites manycases such as the restaurant waiter asking Who is the ham sandwich? (Nunberg,1995, p. 115). Copestake and Briscoe (1992, p. 116) discuss the conversion ofanimal terms to mass nouns. Example (22) is about a substance, not about a cutebunny.

(22) After several lorries had run over the body, there was rabbit splattered allover the road.

14Kay (2005), working in the framework of CxG, also suggests unary constructions.




The authors suggest a lexical rule that maps a count noun onto a mass noun. Thisanalysis is also assumed by Fillmore (1999, p. 114–115). Such coercion can occurwithout any syntactic context: one can answer the question What’s that stuff onthe road? or What are you eating? with the one-word utterance Rabbit. Somecoercion happens to affect the complement structure of a verb, but this is simplya special case of a more general phenomenon that has been analyzed by rules ofsystematic polysemy.

5.3 Valence and Coordination

On the lexical account, the verb paint in (21b), for example, is lexically a 2-argument verb, while the unary branching node immediately dominating it is ef-fectively a 3-argument verb. On the constructional view there is no such predicateseeking three arguments that dominates only the verb. Coordination provides evi-dence for the lexical account.

A generalization about coordination is that two constituents that have compat-ible syntactic properties can be coordinated and that the result of the coordinationis an object that has the syntactic properties of each of the conjuncts. This is re-flected by the Categorial Grammar analysis which assumes the category (X\X)/Xfor the conjunction: The conjunction takes an X to the right, an X to the left andthe result is an X.

For example, in (23a) we have a case of the coordination of two lexical verbs.The coordination know and like behaves like the coordinated simplex verbs: Ittakes a subject and an object. Similarly, two sentences with a missing object arecoordinated in (23b) and the result is a sentence with a missing object.

(23) a. I know and like this record.

b. Bagels, I like and Ellison hates.

The German examples in (24) show that the case requirement of the involved verbshave to be observed. In (24b,c) the coordinated verbs require accusative and dativerespectively and since the case requirements are incompatible with unambigiouslycase marked nouns both of these examples are out.

(24) a. IchI

kenneknow

undand

unterstützesupport

diesenthis

Mann.man.ACC

b. * IchI

kenneknow

undand

helfehelp

diesenthis

Mann.man.ACC

c. * IchI

kenneknow

undand

helfehelp

diesemthis

Mann.man.DAT



5.3 Valence and Coordination 23

Interestingly, it is possible to coordinate basic ditransitive verbs with verbs thathave additional arguments licensed by the lexical rule. (25)–(26) provide someexamples:

(25) a. My sisters just baked and gave me a nutella cupcake with mint choco-late chip ice-cream in the middle and milk chocolate frosting on top.15

b. Randy baked and gave me a loaf of bread!!!16

(26) a. I’ve never had a problem if I want extra sauce or veggie rice, in factthe last time I was in, they offered and made me a special sauce for mycrab rangoon.17

b. She then offered and made me a wonderful espresso — nice.18

c. I always passed and greeted a woman working in her front yard alongthe way, and one day in October, when it was absolutely disgustinglyhot and humid outside, she offered and made me a glass of iced tea.19

d. I offered and made him a coffee, white with one and he set to work.20

Note that in (25) the monotransitive verb precedes the ditransitive verb, while in(26) they appear in the opposite order. These sentences show that both verbs are3-argument verbs at the V 0 level, since they involve V 0 coordination:

(27) a. [baked and gave]V 0 [me]DP [a cake]DP

b. [offered and made]V 0 [me]DP [a special sauce]DP

This is expected under the lexical rule analysis but not the non-lexical construc-tional one.

One might wonder whether these sentences could be instances of Right NodeRaising (RNR) out of coordinated VPs:

(28) a. [ baked ___ ] and [ gave me ___ ] a cake

b. [ offered ___ ] and [ made me ___ ] a special sauce

But there are problems with assuming RNR analyses of all such cases, to theexclusion of V 0 coordination. First, note that under such an analysis the firstverb has been used without a benefactive or recipient object. But the verb offer

15http://bambambambii.tumblr.com/post/809470379. 05.06.2012.16http://marilynmillersstudio.blogspot.com/2012/03/floral-celebration-of-friendships.html.

16.07.2012.17http://www.yelp.com/biz/king-wok-47-sugar-grove 07.07.201218http://www.thespinroom.com.au/?p=102 07.07.201219http://talk.collegeconfidential.com/duke-university/1242028-current-students-what-

distinguishes-duke-2.html 07.07.201220http://dv-photography.blogspot.com/ 07.07.2012


http://bambambambii.tumblr.com/post/809470379

http://marilynmillersstudio.blogspot.com/2012/03/floral-celebration-of-friendships.html

http://www.yelp.com/biz/king-wok-47-sugar-grove

http://www.thespinroom.com.au/?p=102

http://talk.collegeconfidential.com/duke-university/1242028-current-students-what-distinguishes-duke-2.html

http://talk.collegeconfidential.com/duke-university/1242028-current-students-what-distinguishes-duke-2.html

http://dv-photography.blogspot.com/



without the recipient (?She offered a special sauce) is somewhat more awkwardthan the sentences in (26). Also, extraction can be used to distinguish RNR fromV 0 coordination. By way of background, note that extraposed constituents areislands to extraction (Ross, 1974; Bresnan, 1976, 1994):

(29) a. I located a picture of you in the police files.

b. I located __ in the police files a picture of you.

c. I consider arguing with women who ride motorcycles silly.

d. I consider __ silly arguing with women who ride motorcycles.

(30) a. Guess who I located a picture of __ in the police files?

b. * Guess who I located in the police files a picture of __?

c. It’s you who(m) I consider arguing with __ silly.

d. * It’s you who(m) I consider silly arguing with __.

(Examples from Bresnan, 1994, p. 88, examples (49)–(50).) The acceptable ex-traction in (31b) shows that (31a–b) has a non-RNR analysis. We assume thecoordinate V 0 offered and drew simply takes two objects.

(31) a. I offered and drew her a picture of an elephant dancing in a tutu.

b. Guess what I offered and drew her a picture of __?

c. I offered her __, and Bill drew her __, a picture of an elephant dancingin a tutu.

d. * Guess what I offered her and Bill drew her a picture of __?

By contrast, (31c) has only a RNR analysis so extraction is not possible. In addi-tion, it is not possible to do RNR with an unstressed pronoun, eliminating such ananalysis of examples like the following:

(32) Someone make and give me it.21

Finally, a RNR analysis is not available for the German counterpart in (33):

(33) ichI

habhave

ihrher

jetztnow

diesethis

Ladungload

MuffinsMuffins

mitwith

denthe

Herzchenlittle.heart

draufthere.on

gebackenbacked

undand

gegeben.22

given‘I have now baked and given her this load of Muffins with the little hearton top.’

21http://www.penmai.com/forums/vegetarian-recipes/1559-aviyal.html22http://www.musiker-board.de/diverses-ot/35977-die-liebe-637-print.html. 08.06.2012


http://www.musiker-board.de/diverses-ot/35977-die-liebe-637-print.html



Summarizing the coordination argument: If it is the verbs that are coordinated andif the coordinated items must have compatible syntactic properties like valence,this means that gebacken (‘baked’) and gegeben (‘given’) have the same valenceproperties. This is accounted for in a lexical approach since in the lexical approachthe transitive version of the creation verb bake licenses a ditransitive verb whichcan be coordinated with give. In the phrasal approach however, the verb bakehas two argument roles and is not compatible with the verb give, which has threeargument roles. In the phrasal model, bake can only realize three arguments whenit enters the ditransitive argument structure construction, but in sentences like (33)it is not bake alone that enters the phrasal syntax, but rather the combination ofbaked and given. But the verbs are incompatible as far as the semantic roles areconcerned.

To fix this under the phrasal approach, one could posit a mechanism such thatthe semantic roles that are required for the coordinate phrase baked and givenare shared by each of its conjunct verbs and that they are therefore compatible.But this would amount in saying that there are several verb senses for baked,something that the anti-lexicalists claim to avoid, as discussed in the next section.

The coordination facts illustrate a more general point: the output of a lexicalrule is just a word (an X0), so it has the same syntactic distribution as an underivedword with the same category and valence feature. This important generalizationfollows from the lexical account while on the phrasal view it is at best mysterious.The point can be shown with any of the lexical rules that the anti-lexicalists areso keen to eliminate in favor of phrasal constructions. For example, the active andpassive verbs can be coordinated, as long as they have the same valence properties,as in these Swedish examples:

(34) HanHe

dogdied

ochand

glömde-sforget-PASS

bort.away

‘He died and was forgotten.’

(35) Viwe

beställdeordered

ochand

serverade-sserved-PASS

ena

bragood

CheersCheers

chowderchowder

tillto

attINF

börjastart

med,with

ochand

sedanthen

ena

storbig

hummerlobster

varje.23

each‘We ordered and were served a good Cheers chowder to start with, and thena big lobster each.’

(English works the same way, as shown by the grammatical translation lines, butthe periphrastic form of the passive complicates issues slightly.) In (34) the ac-tive and passive are both intransitive, while in (35) the passive of the ditransitive

23http://www.tripadvisor.se/ShowUserReviews-g40990-d412862-r130540331-Fox_s_Lobster_House-York_Beach_York_Maine.html. 16.07.2012


http://www.tripadvisor.se/ShowUserReviews-g40990-d412862-r130540331-Fox_s_Lobster_House-York_Beach_York_Maine.html

http://www.tripadvisor.se/ShowUserReviews-g40990-d412862-r130540331-Fox_s_Lobster_House-York_Beach_York_Maine.html



verb servera ‘serve’ retains one object, so it is effectively transitive and can becoordinated with the active transitive beställa ‘order’.

Moreover, the English passive verb form, being a participle, can feed a secondlexical rule deriving adjectives from verbs (see Figure 3 above). All categories ofEnglish participles can be converted to adjectives:

(36) a. active present participles (cp. The leaf is falling): the falling leaf

b. active past participles (cp. The leaf has fallen): the fallen leaf

c. passive participles (The toy is being broken (by the child).): the brokentoy

That the derived forms are adjectives, not verbs, is shown by a host of properties,including negative un- prefixation: unbroken means ‘not broken’, just as unkindmeans ‘not kind’, while the un- appearing on verbs indicates, not negation, butaction reversal, as in untie (Bresnan, 2001, Chapter 3). Predicate adjectives pre-serve the subject of predication of the verb and for prenominal adjectives the ruleis simply that the role that would be assigned to the subject goes to the modifiednoun instead (The toy remained (un-)broken.; the broken toy). Being an A0, sucha form can be coordinated with another A0, as in the following:

(37) a. The suspect should be considered [armed and dangerous].

b. any [old, rotting, or broken] toys

In (37b), three adjectives are coordinated, one underived (old), one derived froma present participle (rotting), and one from a passive participle (broken). Suchcoordination is completely mundane on a lexical theory. Each A0 conjunct has avalence feature (in HPSG it would be the SPR feature for predicates or the MODfeature for the prenominal modifiers), which is shared with the mother node of thecoordinate structure. But the point of the phrasal (or ASC) theory is to deny thatwords have such valence features.

The claim that lexical derivation of valence structure is distinct from phrasalcombination is further supported with evidence from deverbal nominalization(Wechsler, 2008b). To derive nouns from verbs, -ing suffixation productively ap-plies to all declinable verbs (the shooting of the prisoner), while morphologicalproductivity is severely limited for various other suffixes such as -(a)tion (*theshootation of the prisoner). So forms such as destruction and distribution mustbe retrieved from memory while -ing nouns such as looting or growing could be(and in the case of rare verbs or neologisms, must be) derived from the verb orthe root through the application of a rule (Zucchi, 1993). This difference explainswhy ing-nominals always retain the argument structure of the cognate verb, whileother forms show some variation. A famous example is the lack of the agent ar-gument for the noun growth versus its retention by the noun growing: *John’s




growth of tomatoes versus John’s growing of tomatoes. However, on Marantz’s(1997) phrasal analysis, a phrasal construction (notated as vP) is responsible forassigning the agent role of growing. For him, none of the words directly selectagents, but -ing forms are permitted to appear in the agent-licensing vP construc-tion, while other nouns cannot. A problem for Marantz is that these two types ofnouns can coordinate and share dependents (examples (38a,b) are from Wechsler,2008b, Section 7):

(38) a. With nothing left after the soldier’s [destruction and looting] of theirhome, they reboarded their coach and set out for the port of Calais.24

b. Anyone with information in relation to the [growing or distribution] ofcannabis is urged to contact Crime Stoppers on 1 800 333 0000.

c. I believe it is time in the USA voting population to have the opportunityto vote on adding an amendment to the Bill of Rights to legalize the[use, growth and selling] of marijuana.25

On the phrasal analysis, the nouns looting, growing, and selling occur in one typeof syntactic environment (namely vP), while forms destruction, distribution, use,and growth occur in a different syntactic environment. This places contradictorydemands on the structure of sentences like (38). As far as we know, neither thisproblem nor the others raised by Wechsler (2008b) have even been addressed byadvocates of the phrasal theory of argument structure.

Consider one last example. In an influential phrasal analysis, Hale and Keyser(1993) derived denominal verbs like to saddle through noun incorporation out of astructure akin to [PUT a saddle ON x]. Again, verbs with this putative derivationroutinely coordinate and share dependents with verbs of other types:

(39) Realizing the dire results of such a capture and that he was the only one toprevent it, he quickly [saddled and mounted] his trusted horse and with agrim determination began a journey that would become legendary.26

As in all of these X0 coordination cases, under the phrasal analysis the two verbsplace contradictory demands on a single phrase structure.

A lexical valence structure is an abstraction or generalization over various oc-currences of the verb in syntactic contexts. To be sure, one key use of that valencestructure is simply to indicate what sort of phrases the verb must (or can) combinewith, and the result of semantic composition; if that were the whole story then thephrasal theory would be viable. But it is not. As it turns out, this lexical valence

24http://www.amazon.com/review/R3IG4M3Q6YYNFT 21.07.1225http://signon.org/sign/constitutional-amendment-2826http://www.jouetthouse.org/index.php?option=com_content&view=article&id=56&Itemid=63

21.07.12


http://www.amazon.com/review/R3IG4M3Q6YYNFT

http://signon.org/sign/constitutional-amendment-28

http://www.jouetthouse.org/index.php?option=com_content&view=article&id=56&Itemid=63



structure, once abstracted, can alternatively be used in other ways: among otherpossibilities, the verb (crucially including its valence structure) can be coordinatedwith other verbs that have a similar valence structure; or it can serve as the inputto lexical rules specifying a new word bearing a systematic relation to the inputword. The coordination and lexical derivation facts follow from the lexical view,while the phrasal theory at best leaves these facts as mysterious and at worst leadsto irreconcilable contradictions for the phrase structure.

5.4 Valence and Derivational Morphology

Goldberg and Jackendoff (2004), Alsina (1996), and Asudeh, Dalrymple andToivonen (2008) suggest analyzing resultative constructions and/or caused mo-tion constructions as phrasal constructions. As was argued in Müller, 2006 this isincompatible with the assumption of Lexical Integrity, that is, that word formationhappens before syntax (Bresnan and Mchombo, 1995). Let us consider a concreteexample, such as (40):

(40) a. Erhe

tanztdances

diethe

Schuheshoes

blutigbloody

/ ininto

Stücke.pieces

b. diethe

ininto

Stückepieces

/ blutigbloody

getanztendanced

Schuheshoes

c. * diethe

getanztendanced

Schuheshoes

The shoes are not a semantic argument of tanzt. Nevertheless the NP that is re-alized as accusative NP in (40a) is the element the adjectival participle in (40b)predicates over. Adjectival participles like the one in (40b) are derived from a pas-sive participle of a verb that governs an accusative object. If the accusative objectis licensed phrasally by configurations like the one in (40a) it cannot be explainedwhy the participle getanzte can be formed despite the absence of an accusativeobject. See Müller, 2006, Section 5 for further examples of the interaction ofresultatives and morphology. The conclusion, which was drawn in the late 70sand early 80s by Dowty (1978, p. 412) and Bresnan (1982c, p. 21), is that phe-nomena that feed morphology should be treated lexically. The natural analysis inframeworks like HPSG, CG, CxG, and LFG is therefore one that assumes a lexicalrule for the licensing of resultative constructions. See Verspoor, 1997; Wechsler,1997, Wechsler and Noh, 2001, Müller, 2002a, Kay, 2005, and Simpson, 1983 forlexical proposals in the some of these frameworks.

This argument is similar to the one that was discussed in conenction with theGPSG representation of valence in Section 4.2: morphological processes have tobe able to see the valence of the element they attach to. This is not the case ifarguments are introduced by phrasal configurations after the morphology level.



5.5 Simplicity and polysemy 29

5.5 Simplicity and polysemy

Much of the intuitive appeal of the plugging approach stems from its apparentsimplicity relative to the use of lexical rules. But the claim to greater simplicityfor Construction Grammar is based on misunderstandings of both lexical rulesand Construction Grammar (specifically of Goldberg’s (1995, 2006) version). Itdraws the distinction in the wrong place and misses the real differences betweenthese approaches. This argument from simplicity is often repeated and so it isimportant to understand why it is incorrect.

Tomasello (2003) presents the argument as follows. Discussing first the lexicalrules approach, Tomasello (2003, p. 160) writes that

One implication of this view is that a verb must have listed in thelexicon a different meaning for virtually every different constructionin which it participates [. . . ]. For example, while the prototypicalmeaning of cough involves only one participant, the cougher, we maysay such things as He coughed her his cold, in which there are threecore participants. In the lexical rules approach, in order to producethis utterance the child’s lexicon must have as an entry a ditransitivemeaning for the verb cough. (Tomasello, 2003, p. 160)

Tomasello (2003, p. 160) then contrasts a Construction Grammar approach, cit-ing Fillmore et al. (1988), Goldberg (1995), and Croft (2001). He concludes asfollows:

The main point is that if we grant that constructions may have mean-ing of their own, in relative independence of the lexical items in-volved, then we do not need to populate the lexicon with all kindsof implausible meanings for each of the verbs we use in everydaylife. The construction grammar approach in which constructions havemeanings is therefore both much simpler and much more plausiblethan the lexical rules approach. (Tomasello, 2003, p. 161)

This reflects a misunderstanding of lexical rules, as they are normally understood.There is no implausible sense populating the lexicon. The lexical rule approachto He coughed her his cold states that when the word coughed appears with twoobjects, the whole complex has a certain meaning. See Müller (2006, p. 876).

The simplicity argument also relies on a misunderstanding of a theory he ad-vocates, namely the theory due to Goldberg (1995, 2006). For his argument to gothrough, Tomasello must tacitly assume that verbs can combine freely with con-structions, that is, that the grammar does not place extrinsic constraints on suchcombinations. If it is necessary to also stipulate which verbs can appear in which




constructions then the claim to greater simplicity collapses: each variant lexicalitem with its ‘implausible meaning’ under the lexical rule approach correspondsto a verb-plus-construction combination under the phrasal approach.

Passages such as the following may suggest that verbs and constructions areassumed to combine freely:27

Constructions are combined freely to form actual expressions as longas they can be construed as not being in conflict (invoking the notionof construal is intended to allow for processes of accommodation orcoercion). (Goldberg, 2006, p. 22)

Allowing constructions to combine freely as long as there are no con-flicts, allows for the infinitely creative potential of language. [. . . ]That is, a speaker is free to creatively combine constructions as longas constructions exist in the language that can be combined suitablyto categorize the target message, given that there is no conflict amongthe constructions. (Goldberg, 2006, p. 22)

But in fact Goldberg does not assume free combination, but rather that a verbis ‘conventionally associated with a construction’ (Goldberg, 1995, p. 50): verbsspecify their participant roles and which of those are obligatory direct arguments(profiled, in Goldberg’s terminology; see Section 3). In fact Goldberg herself(2006, p. 211) argues against Borer’s putative assumption of free combination2003 on the grounds that Borer is unable to account for the difference betweendine (intransitive), eat (optionally transitive), and devour (obligatorily transitive).28

Indeed, it is well-established that not all realizational patterns are determined bythe meaning of the verb. In addition to lexically specified transitivity, preposi-tion and oblique case selection is sometimes loosely semantically motivated butnonetheless arbitrarily associated with certain verbs: rely on/*in me versus trustin/*on me, and so on. Despite Tomasello’s comment above, Construction Gram-mar is no simpler than the lexical rules.

The resultative construction is often used to illustrate the simplicity argument.For example, Goldberg (1995, Chapter 7) assumes that the same lexical item forthe verb sneeze is used in (41a) and (41b). It is simply inserted into differentconstructions:

(41) a. He sneezed.

27The context of these quotes makes clear that ‘combining constructions’ includes combininga verb with an argument structure construction, since the verb itself is considered a construction.See Goldberg (2006, p. 21, ex. (2)).

28Goldberg’s critique cites a 2001 presentation by Borer with the same title as Borer (2003).See Section 6.3 for more discussion of Borer’s theory.



31

b. He sneezed the napkin off the table.

The meaning of (41a) corresponds more or less to the verb meaning, since theverb is used in the Intransitive Construction. But the Caused-Motion Construc-tion in (41b) contributes additional semantic information concerning the causa-tion and movement: His sneezing caused the napkin to move off the table. sneezeis plugged into the Caused Motion Construction, which licences the subject ofsneeze and additionally provides two slots: one for the theme (napkin) and onefor the goal (off the table).

In a nuanced comparison of the two approaches, Goldberg (1995, p. 139ff)considers again the added recipient argument in Mary kicked Joe the ball, wherekick is lexically a 2-place verb. She notes that on the constructional view, ‘thecomposite fused structure involving both verb and construction is stored in mem-ory’. Since this is recognized as a composite structure, the verb itself retains itsoriginal meaning as a 2-place verb, so that ‘we avoid implausible verb sensessuch as “to cause to receive by kicking”’; the idea seems to be that the lexical ap-proach, in contrast, must countenance such implausible verb senses since a lexicalrule adds a third argument.

But the lexical and constructional approaches are actually indistinguishable onthis point. The lexical rule does not produce a verb with the ‘implausible sense’in (42a), but rather the one in (42b):

(42) a. cause-to-receive-by-kicking(x, y, z)

b. cause(kick(x, y),(receive(y,z))

In short, the same sort of ‘the composite fused structure’ is assumed under eitherview. The particular conception of lexical rules as unary branching structures ac-tually encodes this in the phrase structure, but decomposition is assumed underother lexical approaches as well. With respect to the semantic structure, the num-ber and plausibility of senses, and the polyadicity of the semantic relations, thetwo theories are identical. They differ in one respect: on the lexical theory, the de-rived valence structure, which includes places for three participants, is associatedwith the phonological string kicked. We saw evidence for that claim in Section 5.3above.

6 Radical Underspecification: The End of Argument

Structure?

In the last section we examined proposals that assume that verbs come with certainargument roles and are inserted into prespecified structures that may contributeadditional arguments. While we showed that this is not without problems, there



32 6 Radical Underspecification: The End of Argument Structure?

are even more radical proposals that the construction adds all agent arguments, oreven all arguments. The notion that the agent argument should be severed from itsverbs is put forth by Marantz (1984, 1997), Kratzer (1996), Embick (2004) andothers. Others suggest that no arguments are selected by the verb. Borer (2003)calls such proposals exoskeletal since the structure of the clause is not determinedby the predicate, that is, the verb does not project an inner ‘skeleton’ of the clause.Counter to such proposals are endoskeletal approaches, in which the structure ofthe clause is determined by the predicate, that is, lexical proposals. The radicalexoskeletal proposals are mainly proposed in Mainstream Generative Grammar(Borer, 1994, 2003, 2005; Schein, 1993; Hale and Keyser, 1997; Lohndal, 2012)but can also be found in HPSG (Haugereid, 2009). We will not discuss theseproposals in detail here, but we review the main issues insofar as they relate to thequestion of lexical argument structure. We conclude that the available empiricalevidence favors the lexical argument structure approach over such alternatives.

Davidson (1967) argued for an event variable in the logical form of actionsentences (43a). Dowty (1989) coined the term Neodavidsonian for the variantin (43b), in which the verb translates to a property of events, and the subject andcomplement dependents are translated as arguments of secondary predicates suchas agent and theme. (Dowty (1989) called the system in 43a an ordered argumentsystem.) Kratzer (1996) further noted the possibility of mixed accounts such as(43c), in which the agent (subject) argument is severed from the kill relation, butthe theme (object) remains an argument of the kill relation.

(43) a. kill: λyλx∃e[kill(e, x, y)] (Davidsonian)

b. kill: λyλx∃e[kill(e)∧ agent(e, x)∧ theme(e, y)](Neodavidsonian)

c. kill: λyλx∃e[kill(e, y)∧ agent(e, x)] (mixed)

Kratzer (1996) observed that a distinction between Davidsonian, Neodavidsonianand mixed can be made either ‘in the syntax’ or ‘in the conceptual structure’(Kratzer, 1996, p. 110-1). For example, on a lexical approach of the sort weadvocate here, any of the three alternatives in 43 could be posited as the semanticcontent of the verb kill. A lexical entry for kill on the mixed model appears in 44.

(44)

PHON 〈 kill 〉

ARG-ST⟨

NP x , NP y

⟩

CONTENT kill(e, y)∧ agent(e, x)

In other words, the lexical approach is neutral on this question of the ‘conceptualstructure’ of eventualities, as noted already in a different connection in Section5.5. For that reason, certain semantic arguments for the Neodavidsonian approach,



6.1 Little v and idiom asymmetries 33

such as those put forth by Schein (1993, Chapter 4) and Lohndal (2012), do notdirectly bear upon the issue of lexicalism, as far as we can tell.

But Kratzer (1996), among others, has gone further and argued for an accountthat is neodavidsonian (or rather, mixed) ‘in the syntax’. Kratzer’s claim is that theverb specifies only the internal argument(s), as in (45a) or (45b), while the agent(external argument) role is assigned by the phrasal structure. On the ‘neodavidso-nian in the syntax’ view, the lexical representation of the verb has no argumentsat all, except the event variable, as shown in as in (45c).

(45) a. kill: λy∃e[kill(e, y)] (agent is severed)

b. kill: λy∃e[kill(e)∧theme(e, y)] (agent is severed)

c. kill: ∃e[kill(e, y))] (all arguments severed)

On such accounts, the remaining dependents of the verb receive their semanticroles from silent secondary predicates, which are usually assumed to occupy thepositions of functional heads in the phrase structure. A standard term for theagent-assigning silent predicate is ‘little v’. These extra-lexical dependents are theanalogues of the ones contributed by the constructions in Construction Grammar.

In the following subsections we address arguments that have put forth in favorof the ‘little v’ hypothesis, from idiom asymmetries (Section 6.1) and deverbalnominals (Section 6.2). We argue that the evidence actually favors the lexicalview. Then we turn to problems for exoskeletal approaches, from idiosyncraticsyntactic selection (Section 6.3) and expletives (Section 6.4). We conclude with alook at the exoskeletal alternative (Section 6.5), and a summary (Section 6.6)

6.1 Little v and idiom asymmetries

Marantz (1984) and Kratzer (1996) argued for severing the agent from the ar-gument structure as in 45a, on the basis of putative idiom asymmetries. Marantz(1984) observed that while English has many idioms and specialized meanings forverbs in which the internal argument is the fixed part of the idiom and the externalargument is free, the reverse situation is considerably rarer. To put it differently,the nature of the role played by the subject argument often depends on the filler ofthe object position, but not vice versa. To take Kratzer’s examples (Kratzer, 1996,p. 114):

(46) a. kill a cockroach

b. kill a conversation

c. kill an evening watching TV

d. kill a bottle (i.e. empty it)

e. kill an audience (i.e., wow them)




On the other hand, one does not often find special meanings of a verb associ-ated with the choice of subject, leaving the object position open (examples fromMarantz (1984, p. 26)):

(47) a. Harry killed NP.

b. Everyone is always killing NP.

c. The drunk refused to kill NP.

d. Silence certainly can kill NP.

Kratzer observes that a mixed representation of kill as in (48a) allows us to specifyvarying meanings that depend upon its sole NP argument.

(48) a. kill: λy∃e[kill(e, y)]

b. If a is a time interval, then kill(a,e) = truth if e is an event of wasting aIf a is animate, then kill(a,e) = truth if e is an event in which a dies. . . etc.

On the polyadic (Davidsonian) theory, the meaning could similarly be made to de-pend upon the filler of the agent role. On the polyadic view, ‘there is no technicalobstacle’ (Kratzer, 1996, p. 116) to conditions like those in 48b, except reversed,so that it is the filler of the agent role instead of the theme role that affects themeaning. But, she writes, this could not be done if the agent is not an argumentof the verb. According to Kratzer, the agent-severed representation (such as 48a)disallows similar constraints on the meaning that depend upon the agent, therebycapturing the idiom asymmetry.

But as noted by Wechsler (2005), ‘there is no technical obstacle’ to specifyingagent-dependent meanings even if the Agent has been severed from the verb asKratzer proposes. It is true that there is no variable for the agent in 48a. But thereis an event variable e, and the language user must be able to identify the agentof e in order to interpret the sentence. So one could replace the variable a with‘the agent of e’ in the expressions in 48b, and thereby create verbs that violate theidiom asymmetry.

While this may seem to be a narrow technical or even pedantic point, it isnonetheless crucial. For Kratzer’s argument to go through, it requires an addi-tional assumption: that modulations in the meaning of a polysemous verb canonly depend upon arguments of the relation denoted by that verb, and not onother participants in the event. But under that additional assumption, it makes nodifference whether the agent is severed from the lexical entry or not. For exam-ple, consider the following (mixed) neodavidsonian representation of the semanticcontent in the lexical entry of kill:

(49) kill: λyλx∃e[kill(e, y) ∧ agent(e, x)]



6.2 Deverbal nominals 35

Assuming that sense modulations can only be affected by arguments of the kill(e,y)relation, we derive the idiom asymmetry, even if (49) is the lexical entry for kill.

Moreover, recasting Kratzer’s account in lexicalist terms allows for verbs tovary. This is an important advantage, because the putative asymmetry is only atendency. Following are examples in which the subject is a fixed part of the idiomand there are open slots for non-subjects:

(50) a. A little bird told X that S.‘X heard the rumor that S’) (Nunberg et al., 1994, p. 526)

b. The cat’s got x’s tongue.‘X cannot speak.’ (Bresnan, 1982a, p. 349–350)

c. What’s eating x?‘Why is X so galled?’ (Bresnan, 1982a, p. 349–350)

Further data and discussion of subject idioms in English and German can be foundin the Appendix below.

The tendency towards a subject-object asymmetry plausibly has an indepen-dent explanation. Nunberg et al. (1994) argue that the subject-object asymmetryis a side-effect of an animacy asymmetry. The open positions of idioms tend to beanimate while the fixed positions tend to be inanimate. Nunberg et al. (1994) de-rive these animacy generalizations from the figurative and proverbial nature of themetaphorical transfers that give rise to idioms. If there is an independent explana-tion for this tendency, then a lexicalist grammar successfully encodes those pat-terns, perhaps with a mixed neodavidsonian lexical decomposition, as explainedabove (see Wechsler (2005) for such a lexical account of the verbs buy and sell).But the ‘little v’ hypothesis rigidly predicts this asymmetry for all agentive verbs,and that prediction is not borne out.

6.2 Deverbal nominals

An influential argument against lexical argument structure involves English de-verbal nominals and the causative alternation. It originates from a mention inChomsky (1970), and is developed in detail by Marantz (1997); see also Peset-sky (1996) and Harley and Noyer (2000). The argument is often repeated, but itturns out that the empirical basis of the argument is incorrect, and the actual factspoint in the opposite direction, in favor of lexical argument structure (Wechsler,2008a,b).

Certain English causative alternation verbs allow optional omission of theagent argument (51), while the cognate nominal disallows expression of the agent(52):

(51) a. that John grows tomatoes




b. that tomatoes grow

(52) a. *John’s growth of tomatoes

b. the tomatoes’ growth, the growth of the tomatoes

In contrast, nominals derived from obligatorily transitive verbs such as destroyallow expression of the agent, as shown in (54a):

(53) a. that the army destroyed the city

b. *that the city destroyed

(54) a. the army’s destruction of the city

b. the city’s destruction

Following a suggestion by Chomsky (1970), Marantz (1997) argued on the basisof these data that the agent role is lacking from lexical entries. In verbal projec-tions (51) and (53) the agent role is assigned in the syntax by little v. Nominalprojections like (52) and (54) lack little v. Instead, pragmatics takes over to de-termine which agents can be expressed by the possessive phrase: the possessivecan express ‘the sort of agent implied by an event with an external rather than aninternal cause’ because only the former can be ‘easily reconstructed’ (quoted fromMarantz (1997); see also Harley and Noyer (2000)). The destruction of a city hasa cause external to the city, while the growth of tomatoes is internally caused bythe tomatoes themselves (Smith, 1970). Marantz points out that this explanationis unavailable if the noun is derived from a verb with an argument structure spec-ifying its agent, since the deverbal nominal would inherit the agent of a causativealternation verb.

The empirical basis for this argument is the putative mismatch between theallowability of agent arguments, across some verb-noun cognate pairs: e.g. growallows the agent but growth does not. But it turns out that the grow/growth patternis rare. Most deverbal nominals precisely parallel the cognate verb: if the verbhas an agent, so does the noun. Moreover, there is a ready explanation for theexceptional cases that exhibit the grow/growth pattern (Wechsler, 2008b). Firstconsider non-alternating theme-only intransitives (‘unaccusatives’), as in (55) andnon-alternating transitives as in (56). The pattern is clear: if the verb is agentless,then so is the noun:

(55) arriv(al), disappear(ance), fall, etc.:

a. A letter arrived.

b. the arrival of the letter

c. * The mailman arrived a letter.

d. * the mailman’s arrival of the letter



6.2 Deverbal nominals 37

(56) destroy/destruction, construct(ion), creat(ion), assign(ment) etc.:

The army is destroying the city.

a.b. the army’s destruction of the city

This favors the view that the noun inherits the lexical argument structure of theverb. For the anti-lexicalist, the badness of (55c) and (55d), respectively, wouldhave to receive independent explanations. For example, on Harley and Noyer’s2000 proposal, (55c) is disallowed because a feature of the root ARRIVE pre-vents it from appearing in the context of v, but the badness of (55d) is insteadruled out because the cause of an event of arrival cannot be easily reconstructedfrom world knowledge. This exact duplication in two separate components of thelinguistic system would have to be replicated across all non-alternating intransi-tive and transitive verbs, a situation that is highly implausible.

Turning to causative alternation verbs, Marantz’s argument is based on the im-plicit generalization that noun cognates of causative alternation verbs (typically)lack the agent argument. But apart from the one example of grow/growth, there donot seem to be any clear cases of this pattern. Besides grow(th), Chomsky 1970,examples 7c and 8c cited two experiencer predicates, amuse and interest: Johnamused (interested) the children with his stories versus *John’s amusement (in-terest) of the children with his stories. But this was later shown by Rappaport 1983and Dowty 1989 to have an independent aspectual explanation. Deverbal expe-riencer nouns like amusement and interest typically denote a mental state, wherethe corresponding verb denotes an event in which such a mental state comes aboutor is caused. These result nominals lack not only the agent but all the eventive ar-guments of the verb, because they do not refer to events. Exactly to the extentthat such nouns can be construed as representing events, expression of the agentbecomes acceptable.

In a response to Chomsky 1970, Carlota Smith (1972) surveyed Webster’sdictionary and found no support for Chomsky’s claim that deverbal nominals donot inherit agent arguments from causative alternation verbs. She listed manycounterexamples, including ‘explode, divide, accelerate, expand, repeat, neutral-ize, conclude, unify, and so on at length.’ (Smith, 1972, p. 137). Harley andNoyer (2000) also noted many so-called ‘exceptions’: explode, accumulate, sep-arate, unify, disperse, transform, dissolve/dissolution, detach(ment), disengage-(ment). . . . The simple fact is that these are not exceptions because there is nogeneralization to which they can be exceptions. These long lists of verbs repre-sent the norm, especially for suffix-derived nominals (in -tion, -ment, etc.). Manyzero-derived nominals from alternating verbs also allow the agent, such as change,release, and use: My constant change of mentors from 1992-1997. The frequentrelease of the prisoners by the governor. The frequent use of sharp tools by un-derage children. (examples from Borer (2003, fn. 13)). Pesetsky (1996, p. 79, ex.




231) assigns a star to the thief’s return of the money, but it is acceptable to manyspeakers, the Oxford English Dictionary lists a transitive sense for the noun return(definition 11a), and corpus examples like her return of the spoils are not hard tofind.

Like the experiencer nouns mentioned above, many zero-derived nominalslack event readings. Some reject all the arguments of the corresponding eventiveverb, not just the agent: *the freeze of the water, *the break of the window, and soon. In the judgment of the second author, his drop of the ball is slightly odd, butthe drop of the ball has exactly the same degree of oddness. The locution a dropin temperature matches the verbal one The temperature dropped, and both verbaland nominal forms disallow the agent: *The storm dropped the temperature. *thestorm’s drop of the temperature. In short, the facts seem to point in exactly the op-posite direction from what has been assumed in this oft-repeated argument againstlexical valence. Apart from the one isolated case of grow(th), event-denoting de-verbal nominals match their cognate verbs in their argument patterns.

Turning to grow(th), we find a simple explanation for its unusual behavior.When the noun growth entered the English language, causative (transitive) growdid not exist. The OED provides these dates of the earliest attestations of growand growth:

(57) a. intransitive grow: c725 ‘be verdant’ ... ‘increase’ (intransitive)

b. the noun growth: 1587 ‘increase’ (intransitive)

c. transitive grow: 1774 ‘cultivate (crops)’

Thus growth entered the language at a time when transitive grow did not exist.The argument structure and meaning were inherited by the noun from its sourceverb, and then preserved into present-day English. This makes perfect sense if,as we claim, words have predicate argument structures. Nominalization by -thsuffixation is not productive in English, so growth is listed in the lexicon. Toexplain why growth lacks the agent we need only assume that a lexical entry’spredicate argument structure dictates whether it takes an agent argument or not.So even this one word provides evidence for lexical argument structure.

6.3 Idiosyncratic Syntactic Selections

As was mentioned at the beginning of this section, proponents of so-called neo-constructivist approaches assume that roots are stored in the lexicon and connectedto encyclopedic knowledge that helps to determine which arguments may be orhave to be present. The arguments are licenced by functional projections that maycontribute meaning to the core meaning contributed by the root or in Haugereid’s



6.3 Idiosyncratic Syntactic Selections 39

proposal by binary branching ID schemata that licence an argument that fills oneof five argument roles.

The notion of lexical valence structure immediately explains why the argu-ment realization patterns are strongly correlated with the particular lexical headsselecting those arguments. It is not sufficient to have general lexical items with-out valency information and let the syntax and world knowledge decide aboutargument realizations, because not all realizational patterns are determined bythe meaning. The form of the preposition of a prepositional object is sometimesloosely semantically motivated but in other cases arbitrary. For example, the va-lence structure of the English verb depend captures the fact that it selects an on-PPto express one of its semantic arguments:

(58) a. John depends on Mary. (counts, relies, etc.)

b. John trusts (*on) Mary.

c.

PHON 〈 depend 〉

ARG-ST⟨

NP x , PP[on] y

⟩

CONTENT depend(x,y)

Such idiosyncratic lexical selection is utterly pervasive in human language. Theverb or other predicator often determines the choice between direct and obliquemorphology, and for obliques, it determines the choice of adposition or obliquecase. In some languages such as Icelandic even the subject case can be selectedby the verb (Zaenen et al., 1985).

Compare the German translation of the English example in (59a):

(59) a. I am waiting for my man.

b. IchI

wartewait

aufon

meinenmy

Mann.man.ACC

While the English verb wait requires the preposition for (German für) the prepo-sition auf (‘on’) with accusative is used. A learner has to acquire that warten hasto be used with auf + accusative and not with other prepositions or other case.

Similarly it is often impossible to find semantic motivation for case. For in-stance there is a tendency to avoided the genitive in German in favour of the dative:Instead of the genitive as in (60a) one also finds examples with the dative as in(60b):

(60) a. dassthat

derthe

Opfervictims.GEN

gedachtremembered

werdewas

‘that the victims would be remembered’




b. daßthat

auchalso

hierhere

denthe

Opfernvictims.DAT

desof.the

Faschismusfascism

gedachtremembered

werdewas

[. . . ]29

‘that the victims of fascism would be remembered here too’

Verbs like helfen (‘to help’) and unterstützen (‘to support’) are rather similar se-mantically, but govern different cases:

(61) a. Erhe

unterstütztsupports

ihn.him.ACC

b. Erhe

hilfthelps

ihm.him.DAT

In order to avoid that the verb helfen appears in the syntactic environment thatlicences (61a) and that the verb unterstützen appears in the construction that li-cences (61b), one has to specify the case that the respective verbs require in thelexical items of the verbs.30

Pollard and Sag (1987, p. 126) discuss a similar example with the two verbstreffen and begegnen (‘to meet’):

(62) a. Erhe.NOM

trafmet

denthe

Mann.man.ACC

b. Erhe.NOM

begegnetemet

demthe

Mann.man.DAT

Without any semantic motivation one verb takes an accusative object and the otherone takes a dative.

Proponents of neo-constructionist approaches either admit that they cannot ac-count for such contrasts, or make proposals that are difficult to distinguish fromlexical valence structures (see Section 6.5 below). In a phrasal constructionistapproach one would have to assume phrasal patterns into which warten can beinserted and a representation for warten, in which it is represented that wartentakes a prepositional object with auf and an accusative NP (see Kroch and Joshi,1985, Section 5.2 for such a proposal in the framework of TAG). Since generaliza-tions regarding verbs with such valence representations can be found one wouldbe forced to have two inheritance hierarchies: one for lexical entries with theirvalency properties and another one for specific phrasal patterns that are neededfor the specific constructions in which these lexical items can be used.

29Frankfurter Rundschau, 07.11.1997, S. 6.30Or at least mark the fact that unterstützen takes an object with the default case for objects and

helfen takes a dative object in German. See Haider, 1985, Heinz and Matiasek, 1994, and Müller,2001 for structural and lexical case.



6.3 Idiosyncratic Syntactic Selections 41

A radical variant of the plugging approach is suggested by Haugereid (2009).Haugereid (pages 12–13) assumes that the syntax combines a verb with an arbi-trary combination of a subset of five different argument roles. Which argumentscan be combined with a verb is not restricted by the lexical item of the verb.31

Such radically underspecified proposals cannot explain the infelicity of (63):

(63) # Ich verspreche dir, das nicht zu borgen. (Max, 4;9)Intended: ‘I promise to you not to lend it to somebody.’

(63) was uttered in a situation where it was discussed whether a toy squirrel couldbe taken to the kindergarten or not. The promise was to not lend the toy to anybodyelse. The German verb borgen is ambiguous in an interesting way: it has thetwo translations borrow and lend, which basically are two different perspectiveson the same event. See Kunze, 1991, 1993 for an extensive discussion of verbsof exchange of possession. Similar pairs are give and take, buy and sell. Theinteresting fact to note about borgen is that the dative object is obligatory with theborrow′ reading:

(64) a. IchI

borgelend

ihmhim

dasthe

Eichhörnchen.squirrel

‘I lend the squirrel to him.’

b. IchI

borgeborrow

(mir)me

dasthe

Eichhörnchen.squirrel

‘I borrow the squirrel.’

If we omit it, we get the lend′ reading. So, instead of (63), Max should haveuttered (65a) or (65b):

(65) a. IchI

versprechepromise

dir,you

dasit

niemandemnobody

zuto

borgen.lend

b. IchI

versprechepromise

dir,you

dasit

nichtnot

zuto

verborgen.lend.out

It follows that all theories have to have a place where it is fixed that certain ar-guments are necessary for a certain verb meaning or a certain perspective on anevent, respectively.

Webelhuth (1995, p. 34) pointed out the nice minimal triplet dine, devour, andeat. dine is obligatorily intransitive, devour is transitive, and eat can be used bothintransitively and transitively. Here, we have another example of different valencyframes, without there being any possibility to reduce this to a semantic contrast

31Haugereid has the possibility to impose valence restrictions on verbs, but he claims that heuses this possibility just in order to get a more efficient processing of his computer implementation(p. 13).




with regard to the core meaning of the involved predicates: all three involve aneating frame.

The problem of the argument realization in the triplet dine, devour, and eatand other examples by Levin and Rappaport Hovav (2005) that show that certainarguments are obligatory is sometimes noted in the literature, but they are simplyignored. For instance Lohndal (2012) writes:

An unanswered question on this story is how we ensure that the func-tional heads occur together with the relevant lexical items or roots.This is a general problem for the view that Case is assigned by func-tional heads, and I do not have anything to say about this issue here.(Lohndal, 2012)

We think that this view is illegitimate at the current state of linguistics. Gettingcase assigment right in simple sentences without vast overgeneration of ill-formedword sequences is a minimal requirement for a linguistic theory that is asked tobe taken seriously.

6.4 Expletives

A final example for the irreducibility of valence to semantics are verbs that selectfor expletives and reflexive arguments of inherently reflexive verbs in German:

(66) a. weilbecause

esit

regnetrains

b. weilbecause

(es)EXPL

mirme

(vorbefore

derthe

Prüfung)exam

grautdreads

c. weilbecause

erhe

esEXPL

bisuntil

zumto.the

Professorprofessor

bringtbrings

‘because he made it to professor’

d. weilbecause

esEXPL

sichREFL

umaround

denthe

MontagMonday

handelttrades

‘It is about the Monday.’

e. weilbecause

ichI

michmyself

(jetzt)now

erholerecreate

‘because I recreate’

The lexical heads in (66) need to contain information about the expletive sub-jects/objects and/or reflexive pronouns that do not fill semantic roles. Note thatGerman allows for subjectless predicates and hence the presence of expletive sub-jects cannot be claimed to follow from general principles. (66c) is an examplewith an expletive object. Explanations referring to the obligatory presence of a



6.5 Is there an alternative? 43

subject would fail on such examples in any case. Furthermore it has to be ensuredthat erholen is not realized in the [Sbj IntrVerb] construction for intransitive verbsor respective functional categories in a Minimalist setting although the relationerholen′ (recreate′) is a one-place predicate and hence erholen is semanticallycompatible with the construction.

6.5 Is there an alternative?

The question for theories denying the existence of valence structure is what re-places it to explain idiosyncratic lexical selection. In her exoskeletal approach,Borer (2005) explicitly rejects lexical valence structures. But she posits post-syntactic interpretive rules that are difficult to distinguish from them. To explainthe correlation of depend with an on-PP, she posits the following interpretive ruleBorer (2005, Vol. II, p. 29):

(67) MEANING ⇔ π9 + [〈eon〉]

Borer refers to all such cases of idiosyncratic selection as idioms. In rule such asthis one, ‘MEANING is whatever the relevant idiom means.’ (Borer, 2005, Vol.II, p. 27) In this rule, π9 is the ‘phonological index’ of the verb depend and eon

‘corresponds to an open value that must be assigned range by the f-morph on’(ibid), where f-morphs are function words or morphemes. Hence this rule bringstogether much the same information as the lexical valence structure in (58c). Dis-cussing such ‘idiom’ rules, Borer writes

Although by assumption a listeme cannot be associated with any gram-matical properties, one device used in this work has allowed us to getaround the formidable restrictions placed on the grammar by sucha constraint—the formation of idioms. . . . Potentially, then, withinthe system developed here, any syntactic or morphological propertywhich does not reduce directly to some formal computational princi-ple is to be captured by classifying the relevant item as an idiom—apartial representation of a phonological index with some functionalvalue. . . . Such idiomatic specification could be utilized, potentially,not just for arrive and depend on, but also for obligatorily transitiveverbs . . . , for verbs such as put, with their obligatory locative, and forverbs which require a sentential complement.

The reader may object that subcategorization, of sorts, is introducedhere through the back door, with the introduction, in lieu of lexicalsyntactic annotation, of an articulated listed structure, called an id-iom, which accomplishes, de facto, the same task. The objection of




course has some validity, and at the present state of the art, the intro-duction of idioms may represent somewhat of a concession . . . On thepositive side, we note that to the extent that the existence of idiomsis costly, we have attempted to put in place here a system which atleast potentially extricates from the costly component of language allproperties of listemes which are otherwise derivable from the struc-ture.(Borer, 2005, Vol. II, p. 354–5)

Borer goes on to pose various questions for future research, related to constrainingthe class of possible idioms. With regard to that research program, as well as thelast sentence in the above quote, it should be noted that a major focus of lexicalistresearch has been narrowing the class of subcategorization and extricating deriv-able properties from idiosyncratic subcategorization. Those are the functions ofHPSG lexical hierarchies, for example. Whether future research within the ex-oskeletal approach can improve upon the past research in the lexical approachremains to be seen.

6.6 Summary

The discussion in the Sections 14–6.4 shows that the questions which argumentshave to be realized in a sentence cannot be reduced to semantics and world knowl-edge or to general facts about subjects. The consequence is that valence informa-tion has to be connected to lexical items. One therefore has to assume a connectionbetween a lexical item and a certain phrasal configuration as in Croft’s approach(2003) and in LTAG or assume our lexical variant. In a Minimalist setting theright set of features would have to be specified lexically to ensure the presence ofthe right case assigning functional heads. This is basically equivalent to what weare proposing here.

The discussion also showed that information about subjects is to some extentlexeme specific and hence has to be part of the specification of argument structure.This shows that proposals like the one by Goldberg and Jackendoff (2004) thatonly refers to phrasal VP constructions without mentioning the subject are notsufficient. Of course one could represent the fact that a certain head requires asubject lexically and assume a phrasal approach for the VP. This would be a hybridapproach and it seems unclear why one should not assume the lexical approach intotal given that subjects are also involved in valence alternations as in (68):

(68) a. The soup cooked.

b. He cooked the soup.



45

7 Relations between constructions

On the lexical rules approach, word forms are related by lexical rules: a verb stemcan be related to a verb with finite inflection and to a passive verb form; verbs canbe converted to adjectives or nouns; and so on. The lexical argument structureaccompanies the word and can be manipulated by the lexical rule. In this sectionwe consider what can replace such rules within a phrasal or ASC approach.

7.1 Inheritance hierarchies for constructions

We have seen that in a lexical approach one can relate variants of a constructionby lexical rules. In a phrasal approach one would assume that there are allo-constructions for active and passive (for instance Bergen and Chang, 2005; Kall-meyer and Osswald, To appear). If one wants to capture the underlying nature ofpassive one would have to find some way to relate phrasal configurations. This isnecessary since a phrasal approach must otherwise stipulate a great many phrasalconstructions that correspond to observable valency patterns. For instance, onewould assume the ones in (69) for German:

(69) a. Nom V

b. Nom Acc V

c. Nom Dat V

d. Nom Dat Acc V

The passive patterns corresponding to (69) would be:

(70) a. V V-Aux

b. Nom V V-Aux

c. Dat V V-Aux

d. Dat Nom V V-Aux

Researchers working in various frameworks tried to develop inheritance-basedanalyses that could capture the relation between the valency patterns in (69) and(70) (see for instance Kay and Fillmore, 1999, p. 12; Koenig, 1999, Chapter 3;Michaelis and Ruppenhofer, 2001, Chapter 4; Candito, 1996; Clément and Kinyon,2003, p. 188; Koenig, 1999; ?; Kordoni, 2001 for proposals in CxG, TAG, andHPSG), but this enterprise was not successful. For technical problems with theoriginal Berkeley CxG proposal see Müller, 2006, Section 2.4, Section 3.

As was also pointed out in Müller, 2006, Section 4, valency changing pro-cesses in general may not be modeled by inheritance since processes like causativiza-tion can be applied several times. An example is Turkish, which allows doubleand even triple causation (Lewis, 1967, p. 146):



46 7 Relations between constructions

(71) Öl-dür-t-tür-t-‘to cause somebody to cause somebody to kill somebody’

The t and tür is the causative morpheme (-t-/-d- after vowels or sonorants and-tVr-/-dVr after consonants, where V stands for a vowel in vowel harmony). Suchrecursive application of valence-changing rules has been observed in a number oflanguages.

So the only way to capture the generalization with regard to (69) and (70)seems to be to assume GPSG-like meta-rules that relate the constructions in (69)to the ones in (70). If the constructions are lexically linked as in LTAG, the re-spective mapping rules would be lexical rules. For approaches that combine LTAGwith the Goldbergian plugging idea such as the one by Kallmeyer and Osswald(To appear) one would have to have extended families of trees that reflect thepossibility of having additional arguments and would have to make sure that theright morphological form is inserted into the respective trees. The morphologicalrules would be independent of the syntactic structures in which the derived ver-bal lexemes could be used. One would have to assume two independent types ofrules: GPSG-like Meta rules that operate on trees and morphological rules thatoperate on stems and words. We believe that this is an unnecessary complicationand apart from being complicated the morphological rules would not be accept-able as form-meaning pairs in the CxG sense since the aspect of the form thatadditional arguments are required is not captured in these morphological rules. Ifsuch morphological rules were accepted as proper constructions then there wouldnot be any reason left to require that the arguments have to be present in a con-struction in order for it to be recognizable, and hence, the lexical approach wouldbe accepted. Compare the discussion of tot schießen (‘shoot dead’) on page 51.

7.2 Mappings between Different Levels of Representations

Culicover and Jackendoff (2005, Chapter 6.3) suggest that passive should be an-alyzed as one of several possible mappings from the Grammatical Function tierto the surface realization of arguments, that is, an NP in a certain case, with cer-tain agreement properties, or in a certain position. While such analysis that workby mapping elements with different properties onto different representations arecommon in theories like LFG and HPSG (Koenig, 1999; Bouma, Malouf and Sag,2001), a general property of these analyses is that one needs one level of represen-tation per interaction of phenomena (ARG-ST, SEM-ARG, ADD-ARG in Koenig’sproposal, ARG-ST, DEPS, SPR, COMPS in Bouma, Malouf, and Sag’s proposal).This was discussed extensively in Müller, 2008, Section 7.5.2.2 with respect toextensions that would be needed for Koenig’s analysis.

Since Culicover and Jackendoff argue for a phrasal model, we will discuss



7.2 Mappings between Different Levels of Representations 47

their proposal here. The authors write:

we wish to formulate the passive not as an operation that deletes oralters part of the argument structure, but rather as a piece of structurein its own right that can be unified with the other independent piecesof the sentence. The result of the unification is an alternative licensingrelation between syntax and semantics. (Culicover and Jackendoff,2005, p. 203)

They suggest the following representation of the passive:

(72) [GFi > [GF . . . ]]k ⇔ [ . . . Vk + pass . . . (by NPi) . . . ]k

The italicized parts are the normal structure of the sentence and the non-italicizedparts are an overlay on the normal structure, that is additional constraints that haveto hold in passive sentences. GF stands for Grammatical Function. Culicover andJackendoff (2005, p. 204) explicitly avoid names like Subject and Object since thisis crucial for their analysis of the passive to work. They assume that the first GFfollowing a bracket is the subject (p. 195–196) and hence has to be mapped to theappropriate tree position in English. Note that this view on grammatical functionsand obliqueness does not account for subjectless sentences that are possible insome languages, for instance in German.32

Although Culicover and Jackendoff emphasize the similarity between theirapproach and Relational Grammar (Perlmutter, 1983), there is an important dif-ference: In Relational Grammar additional levels (strata) can be stipulated if addi-tional remappings are needed. In Culicover and Jackendoff’s proposal there is noadditional level. This causes problems for the analysis of languages like Lithua-nian (Timberlake, 1982, Section 5), Irish Noonan, 1994, and Turkish (Özkaragöz,1986), which allow for double passivization. We will use Özkaragöz’s Turkishexamples in (73) for illustration (1986, p. 77):

(73) a. Buthis

sato-dachateau-LOC

bog-ul-un-ur.strangle-PASS-PASS-AOR

‘One is strangled (by one) in this chateau.’

b. Buthis

oda-daroom-LOC

döv-ül-ün-ür.hit-PASS-PASS-AOR

‘One is beaten (by one) in this room.’

c. Harp-tewar-LOC

vur-ul-un-ur.shoot-PASS-PASS-AOR

‘One is shot (by one) in war.’

32Of course one could assume empty expletive subjects, as was suggested by Grewendorf (1993,p. 1311), but empty elements and especially those without meaning are generally avoided in theconstructionist literature. See Müller, 2010a, Section 3.4, Section 11.1.1.3 for further discussion.



48 7 Relations between constructions

-In, -n, and -Il are allomorphs of the passive morpheme. According to Özkaragözthe data is best captured by an analysis that assumes that the passive applies to apassivized transitive verb and hence results in an impersonal passive.

Approaches that assume that the personal passive is the unification of a generalstructure with a passive-specific structure will not be able to capture this, sincethey committed to a certain structure too early. The problem for approaches thatstate syntactic structure for the passive is that such a structure, once stated, cannotbe modified. Culicover and Jackendoff’s proposal works in this respect sincethere are no strong constraints in the right-hand side of their constraint in (72).But there is a different problem: When passivization is applied the second time,it has to apply to the innermost bracket, that is, the result of applying (72) shouldbe:

(74) [GFi > [GFj . . . ]]k ⇔ [ . . . Vk + pass . . . (by NPi) . . . (by NPj) . . . ]k

This cannot be done with unification, since unification checks for compatibilityand since the first application of passive was possible it would be possible forthe second time as well. Dots in representations are always dangerous and inthe example at hand one would have to make sure that NPi and NPj are distinct,since the statement in (72) just says there has to be a by PP somewhere. What isneeded instead of unification would be something that takes a GF representationand searches for the outermost bracket and then places a bracket to the left of thenext GF. But this is basically a rule that maps one representation onto another one,just like lexical rules do.

If Culicover and Jackendoff want to stick to a mapping analysis, the only op-tion to analyze the data seems to be to assume an additional level for impersonalpassives from which the mapping to phrase structure is done. In the case of sen-tences like (75), which is a personal passive, the mapping to this level would bethe identity function and in the case of double passivization the correct mappingswould be implemented by two mappings between the three levels that finally resultin a mapping as the one that is seen in (73b).

(75) Arkada-sımfriend-my

buthis

oda-daroom-LOC

döv-ül-dü.hit-PASS-AOR

‘My friend is beaten (by one) in this room.’

Note that the double passivization is also problematic for purely inheritance basedapproaches. What all these approaches can suggest though is that they just stip-ulate three different relations between argument structure and phrase structure:active, passive, double passive. But this misses the fact that (73b) is a furtherpassivization of (75).

In comparison a lexical rule-based approach as the one suggested by Müller(2003a) does not have any problems with double passivization: The first appli-



7.3 Is there an alternative to lexical rules? 49

cation of the passivization lexical rule suppresses the least oblique argument andprovides a lexical item with the argument structure of a personal passive. Thesecond application suppresses the now least oblique argument (the object of theactive clause) and results in an impersonal passive.

7.3 Is there an alternative to lexical rules?

In this section we have presented the attempts to replace lexical rules with meth-ods of relating constructions. We believe those attempts have been unsuccessfulbecause they fail to capture the derivational character of relations between wordforms.

8 Arguments from Language Acquisition

Tomasello (2003) argues for a surface-oriented, pattern-based view on languageacquisition. According to him a child hears sentences like (76) and realizes thatcertain slots can be filled freely (see also Dabrowska, 2001 for analogous propos-als in Cognitive Grammar).

(76) a. Do you want more juice/milk?

b. Mommy is gone.

From such utterances so-called pivot schemata are derived. Such schemata con-tain open slots into which words can be inserted. Examples of schemata that areabstracted from utterances like (76) are shown in (77):

(77) a. more ___ → more juice/milk

b. ___ gone → mommy/juice gone

At this stage (22 months) children do not generalize over such schemata. Theschemata are construction islands and do not have syntax (Tomasello et al., 1997).English children acquire the capability to use novel verbs with a subject and an ob-ject in the SVO order slowly in their third or fourth year of life (Tomasello, 2003,p. 128–129). More abstract syntactic and semantic generalizations only emergein the course of time: after a sufficient amount of encounters of the transitiveconstruction, the child can generalize over the patterns:

(78) a. [S [NP The man/the woman] sees [NP the dog/the rabbit/it]].

b. [S [NP The man/the woman] likes [NP the dog/the rabbit/it]].

c. [S [NP The man/the woman] kicks [NP the dog/the rabbit/it]].



50 8 Arguments from Language Acquisition

According to Tomasello (2003, p. 107) the abstraction of these patterns is [SbjTrVerb Obj], the so-called transitive construction. For Tomasello, children ac-quire the item-specific patterns such as (77)— the ones involving specific verbsare called verb islands— and they eventually also acquire phrasal constructionssuch as the transitive construction.

Verb islands would seem to be perfect precursors to the acquisition of thelexicalist’s valence structures, and indeed Tomasello comments that ‘The lex-ical rules approach would seem to be better adapted to children’s verb islandconstructions and other item-based constructions—which are defined by partic-ular verbs or other words—than to their totally abstract constructions (Tomasello,2003, p. 193). Constructions such as the transitive construction nonetheless carrymeaning, according to Tomasello.

What people crucially never acquire, according to Tomasello, are ‘meaning-less’ phrase structure rules or lexical rules that relate one variant of a word toanother. Constructions seem to take the place of both of these at once, and canplausibly be acquired by children by generalizing directly over the phrasal pat-terns. The constructions at various levels of generality are related to one anotherby inheritance hierarchies (Langacker, 1987; Goldberg, 1995, Chapter 3; Croft,2001, p. 26; Tomasello, 2003, p. 106–107). In language production a number ofsuch constructions combine to form a sentence (Goldberg, 2006, p. 10).

However, this attractive picture depends upon the phrasal view of construc-tions, which has been shown to be untenable. In addition to the problems alreadynoted, a purely pattern-based approach faces difficulties from discontinuities inthe realization of a head and its arguments. For instance, adjuncts can be seri-alized between the subject and the verb. Bergen and Chang (2005, p. 170), whoimplement the phrasal approach, suggest an active-ditransitive construction withthe pattern in (79):

(79) [RefExpr Verb RefExpr RefExpr]

RefExpr stands for referential expression. Their formalization allows a disconti-nuity between the first referential expression and the verb. This makes it possibleto analyze (80a,b), but excludes (80c), since in (80c) the adverb intervenes be-tween verb and the first object:

(80) a. Marry tossed me a drink.

b. Marry happily tossed me a drink.

c. * Marry tossed happily me a drink.

However, by enforcing the adjacency between verb and object the analysis ofcoordinations like (81) becomes impossible.

(81) Marry tossed me a juice and Peter a water.



51

One part of the meaning of this sentence is contributed by the ditransitive con-struction for Marry tossed Peter a water. However, tossed and Peter are discon-tinuous. Similarly, one can construct examples with a discontinuity between thetwo objects of the ditransitive construction:

(82) He showed me and bought for Mary the book that was recommended in theGuardian last week.

me is not adjacent to the book . . . although both phrases are part of the ditransi-tive construction. If one does not use empty elements and dislocation, one cannotmaintain the claim that the items of the ditransitive construction have to be con-tinuous. The point here is that it is not a certain fixed configuration that has tobe acquired but rather the fact that there is a certain dependency between materialin a clause. If material is realized together in a certain syntactic environment, acertain meaning can be observed.

Note also that a purely pattern-based approach is weakened by the existenceof examples like (83):

(83) a. John tried to sleep.

b. John tried to be loved.

Although no argument of sleep is present in the phrase to sleep and neither asubject nor an object is realized in the phrase to be loved, both phrases are rec-ognized as phrases containing an intransitive and a transitive verb, respectively.33

The same applies to arguments that are supposed to be introduced/licensed by aphrasal construction: in (84) the resultative construction is passivized and thenembedded under a control verb, resulting in a situation in which only the resultpredicate and the matrix verb are realized overtly.

(84) Derthe

krankeill

Mannman

wünschtewished

sich,SELF

totdead

geschossenshot

zuto

werden.34

be‘The ill man wanted to be shot dead.’

Of course passivization and control are responsible for these occurrences, but theimportant point here is that arguments can remain unexpressed or implicit andnevertheless a meaning that is usually connected to some overt realization of ar-guments is present (Müller, 2007a, Section 4). So, what has to be acquired by thelanguage learner is that when a result predicate and a main verb are realized to-gether, they contribute the resultative meaning. An additional example that showsthat the NP arguments that are usually realized in active resultative constructionsmay remain implicit are nominalizations like the ones in (85):

33Constructionist theories do not assume empty elements. Of course, in the GB framework thesubject would be realized by an empty element. So it would be in the structure, although invisible.

34Müller, 2007a, p. 387.



52 9 Lexical Items, Licensed Trees, Classes of Trees

(85) a. dannthen

scheintseems

unsus

dasthe

Totschießendead-shooting

mindestensat.least

ebensovielas.much

Spaßfun

zuto

machen35

make‘The shooting dead seems to be as least as much fun’

b. Wirwe

lassenlet

heuttoday

dasthe

Totgeschieße,annoying.repeated.shooting.dead

Weilsince

manone

sowassuch.thing

heuttoday

nichtnot

tut.does

Undand

werwho

einena

Tagday

sichSELF

ausruht,rests

Derthis

schießtshoots

morgentomorrow

doppelttwice

gut.36

good‘We do not shoot anybody today, since one does not do this, and thosewho rest a day shoot twice as good tomorrow.’

9 Lexical Items, Licensed Trees, Classes of Trees

Proponents of the phrasal approach criticize lexical approaches as being inade-quate for capturing language acquisition facts. However, by looking at the prop-erties of lexical approaches carefully, it becomes evident that lexical approachesfulfill the desiderata that are required by the language acquisition facts. As wesaw in the previous section, the fact that certain arguments of a head have to berealized in a certain environment have to be captured by a linguistic theory. Dis-continuities between heads and their arguments have to be allowed. To see thatlexical theories can capture the facts consider Categorial Grammar (Ajdukiewicz,1935; Steedman, 2000). Tomasello’s Transitive Construction [Sbj TrVerb Obj]corresponds to the following lexical item in Categorial Grammar (s\np)/np. Thislexical entry for, let’s say, likes expresses the fact that likes takes an np to its right(marked by the direction of the slash ‘/’) and an np to its left (marked by the di-rection of the slash ‘\’). The lexical item licences structures like the one that isdisplayed as a tree in Figure 4. The combinations are licensed by combinatorialrules that combine a functor with an argument. So all lexical items that are as-signed to the category (s\np)/np can appear in configurations like the one shownin Figure 4. And of course if we observe unknown words in the position of theverb, we can infer that the unknown words have to belong into the same lexicalclass as likes. However, the structure in Figure 4 is not the only one that is pos-sible for items of the category (s\np)/np. For instance an adjunct of the category(s\np)/(s\np) may intervene between the subject and the combination of verb and

35https://www.elitepartner.de/forum/wie-gehen-die-maenner-mit-den-veraenderten-anspruechen-der-frauen-um-26421-6.html. 26.03.0212.

36http://home.arcor.de/finishlast/indexset.html?dontgetmestarted/091201-1.html. 26.03.2012.


https://www.elitepartner.de/forum/wie-gehen-die-maenner-mit-den-veraenderten-

https://www.elitepartner.de/forum/wie-gehen-die-maenner-mit-den-veraenderten-anspruechen-der-frauen-um-26421-6.html

http://home.arcor.de/finishlast/indexset.html?dontgetmestarted/091201-1.html


53

s

np s\np

(s\np)/np np

Kim likes Sandy

Figure 4: Categorial Grammar analysis of Kim likes Sandy

object. This is shown in Figure 5. The adjunct probably takes a VP (s/np) to its

s

np s\np

(s\np)/(s\np) s\np

(s\np)/np np

Kim probably likes Sandy

Figure 5: Categorial Grammar analysis of Kim probably likes Sandy

right and the result of the combination is a VP again.Similarly, lexical items like likes can appear in coordination structures of the

kind discussed above. See Steedman, 1991 for details on coordination.The bracketing in (s\np)/np ensures that the rightmost np in the expression is

combined with the verb first and then the combination with the second np takesplace. This results in the SVO order that can be observed for languages like En-glish. For languages with a freer constituent structure Steedman and Baldridge(2006) suggest a generalized representation. The arguments are represented in aset and for the combination of a head with an argument, one element from this setis chosen. This results in different orders. For a head finial language the lexicalitem of a transitive verb would be s { \np, \np }. See also Hoffman (1995, Sec-tion 3.1) for a similar proposal for Turkish. Such a lexical item stands for trees inwhich the two np arguments precede their head in any order. Such an approach to



54 10 Arguments from Psycholinguistics

constituent order was also suggested by Gunji (1986) in the framework of HPSGand by Fanselow (2001) in the framework of Minimalism.

10 Arguments from Psycholinguistics

Goldberg (1995, Section 1.4.5) uses evidence from psycholinguistic experimentsto argue against lexical approaches that use lexical rules to account for argumentstructure alternations: Carlson and Tanenhaus (1988) showed that sentences withtrue lexical ambiguity like those in (86) and sentences with two verbs with thesame core meaning have different processing times.

(86) a. Bill set the alarm clock onto the shelf.

b. Bill set the alarm clock for six.

(87) a. Bill loaded the truck onto the ship.

b. Bill loaded the truck with bricks.

Errors due to lexical ambiguity cause a bigger increase in processing time thanerrors in the use of the same verb. Experiments showed that there was a biggerdifference in processing times for the sentences in (86) than for the sentencesin (87). The difference in processing times between (87a) and (87b) would beexplained by different preferences for phrasal constructions. In a lexicon-basedapproach one could explain the difference by assuming that one lexical item ismore basic, that is, stored in the mental dictionary and the other is derived fromthe stored one. The application of lexical rules would be time consuming, butsince the lexical items are related, the overall time consumption is smaller thanthe time that is needed to process two unrelated items (Müller, 2002a, p. 405).

Alternatively one could assume that the lexical items for both valency patternsare the result of lexical rule applications. As with the phrasal constructions, thelexical rules would have different preferences. This shows that the lexical ap-proach can explain the experimental results as well, so that they do not force us toprefer phrasal approaches.

Goldberg (1995, p. 18) claims that lexical approaches have to assume two vari-ants of load with different meaning and that this would predict that load alterna-tions would behave like two verbs that really have absolutely different meanings.The experiments discussed above show that such predictions are wrong and hencelexical analyses would be falsified. However, as was shown in Müller, 2010a,Section 11.11.8.2, the argumentation contains two flaws: Let’s assume that theconstruction meaning of the construction that licences (87a) is C1 and the con-struction meaning of the construction that licences (87b) is C2. Under such as-sumptions the semantic contribution of the two lexical items in the lexical analysis



55

would be (88). load(. . . ) is the contribution of the verb that would be assumed inphrasal analyses.

(88) a. load (onto): C1 ∧ load(. . . )b. load (with): C2 ∧ load(. . . )

(88) shows that the lexical items partly share their semantic contribution. Wehence predict that the processing of the dispreferred argument realization of loadis simpler than the dispreferred meaning of set: in the latter case a completelynew verb has to be activated while in the first case parts of the meaning are acti-vated already. (See also Croft, 2003, p. 64–65 for a brief rejection of Goldberg’sinterpretation of the experiment that corresponds to what is said here)

Goldberg (1995, p. 107) argues against lexical rule-based approaches for loca-tive inversions like (89), since according to her such approaches have to assumethat one of the verb forms has to be the more basic form.

(89) a. He loaded hay onto the wagon.

b. He loaded the wagon with hay.

She remarks that this is problematic since we do not have clear intuitions on whatis the basic and what the derived form. She argues that the advantage of phrasalapproaches is that various constructions can be related to each other without ne-cessitating the assumption that one of the constructions is more basic than theother. There are two phrasal patterns and the verb is used in one of the two pat-terns. This criticism can be addressed in two ways: First one could introduce twolexical types (for instance onto-verb and with-verb) into a type hierarchy. The twotypes correspond to two valency frames that are needed for the analysis of (89a)and (89b). These types can have a common supertype (onto-with-verb) which isrelevant for all spray/load verbs. One of the subtypes or the respective lexical itemof the verb is the preferred one. This corresponds to a disjunction in the lexicon,while the phrasal approach assumes a disjunction in the set of constructions.

A variant of this approach is to assume that the lexical description of load justcontains the supertyp, that describes all spray/load verbs. Since Modell TheoreticApproaches assume that all structures that are models of utterances contain onlymaximally specific types (see for instance King, 1999 and Pollard and Sag, 1994,p. 21), it is sufficient to say about verbs like load that they are of type onto-with-verb. Since this type has exactly two subtypes, load has to be either onto-verb orwith-verb in an actual model.37

A second option is to stick with lexical rules and to assume a single represen-tation for the root of a verb that is listed in the lexicon. In addition one assumes

37This analysis does not allow one to specify verb specific preferences for one of the realizationpatterns since the lexicon contains the general type only.



56 11 Arguments from Statistical Distribution of Material

two lexical rules that map this basic lexical item onto other items that can be usedin syntax after being inflected. The two lexical rules can be described by typesthat are part of a type hierarchy and that have a common supertype. This wouldcapture commonalities between the lexical rules. We therefore have the samesituation as with phrasal constructions (two lexical rules vs. two phrasal construc-tions). The only difference is that the action is one level deeper in the lexicalapproach, namely in the lexicon (Müller, 2002a, p. 405–406).

The argumentation with regard to the processing of resultative constructionslike (89c) is parallel:

(90) a. He drinks.

b. He drinks the milk.

c. He drinks the pub empty.

When humans parse a sentence they build up structure incrementally. If one hearsa word that is incompatible with the current hypothesis, the parsing process breaksdown or the current hypothesis is revised. In (90c) the pub does not correspondto the normal transitive use of drink, so the respective hypothesis has to be re-vised. In the phrasal approach the resultative construction would have to be usedinstead of the transitive construction. In the lexical analysis the lexical item thatis licensed by the resultative lexical rule would have to be used rather than the bi-valent one. The building of syntactic structure and lexicon access in general placedifferent demands on our processing capacities. However, when (90c) is parsed,the lexical items for drink are active already, we only have to use a different one.It is currently unclear to us whether psycholinguistic experiments can differentiatebetween the two approaches, but it seems to be unlikely.

11 Arguments from Statistical Distribution of Ma-

terial

In this section, we want to look at arguments from statistics that were claimedto support a phrasal view. We first look at data-oriented parsing, a techniquethat was successfully used by Bod (2009b) to model language acquisition andthen we turn to the collostructional analysis of Stefanowitsch and Gries (2009)and discuss whether this distributional analysis can decide the question whetherargument structure constructions are phrasal or lexical.

11.1 Unsupervised Data-Oriented Parsing

Rens Bod (2009b) demonstrated that a simple statistical procedure can learn quiteelaborated linguistic structures that correspond roughly to what linguists would



11.1 Unsupervised Data-Oriented Parsing 57

assume. In particular he showed that such a technique can learn auxiliary in-version and gets the inversion facts right even for complex examples containingrelative clauses with auxiliaries and even if this type of clause was not in the datathat was used for learning. Chomsky (1971, p. 29–33) has used (and is still us-ing38) auxiliary inversion as his key example of a Poverty of the Stimulus in thelanguage acquisition debate, but Bod has shown that six examples are sufficient toacquire complex auxiliary inversion structures. The examples that are needed donot include the data that Chomsky considered crucial for a language acquisitiondevice that relies on input alone. Bod’s procedure works as follows: An utteranceis partitioned into (binary branching) trees. It is then checked how likely each ofthe subtrees is, that is, it is checked whether an identical subtree occurred in pre-vious utterances. If this is the case, this renders the subtree under considerationmore likely. To take an example, consider the corpus in (91). Figure 6 shows theunlabeled trees for the two sentences.

(91) a. Watch the dog.

b. The dog barks.

X

X

watch the dog

X

X

watch the dog

X

X

the dog barks

X

X

the dog barks

Figure 6: Possible binary branching structures for Watch the dog and The dogbarks.

Some of these trees do not correspond to structures that linguists would assume,but the good news is, that the subtree for the dog appear more often than forinstance watch the and this renders the correct structures for watch the dog andthe dog barks more likely in Bod’s procedure.

If one assumes that language acquisition is based on input alone, the struc-tures that Bod’s procedure extracts from the input must be those that are acquiredby children (part of speech information, meaning, and context are currently not

38Talk in Stuttgart at 24th of March, 2010



58 11 Arguments from Statistical Distribution of Material

included in Bod’s computations, but of course this can be done). By the reverseargument of the poverty of the stimulus the structures that are learned should bethe ones that linguists assume. Since Bod did not have sufficient data to do thecomputations with flat structures and arbitrary branchings, he arbitrarily restrictedthe system to binary branching structures (p. 760). This means that his experi-ments do not answer the question if rules should licence flat structures or binarybranching ones. But we probably can expect interesting results in the future.

What distributional analyses can not determine is where the meaning is rep-resented in a tree. Bod (2009a, p. 132) claims that his procedure is a testablerealization of CxG in Goldberg’s sense, but the trees that he constructs do not helpdeciding between phrasal and lexical analyses or analyses that involve empty ele-ments. The alternative proposals for the analysis of (92) are depicted in Figure 7.

(92) [dass]that

erhe

denthe

Teichpond

leerempty

fischtfishes

‘that he fishes the pond empty’

X

X X

X X

X X

er ihn leer fischt

X

X X

X X

X X

X

er ihn leer fischt

X

X X

X X

X X

X X

er ihn leer fischt _

Figure 7: Three possible analyses for resultative constructions: Holistic construc-tion, lexical rule, empty head

The first picture stands for a complex construction that contributes the resultativemeaning as a whole. The second picture stands for the analysis with a lexical rule,and the third one for an analysis with an empty head. Empty heads are often thechoice in mainstream generative grammar, but as shown in Müller, 2010a, Sec-tion 11.10, some of them can be converted into lexical rules by known techniques



11.2 Collostructions 59

of grammar conversion (Bar-Hillel, Perles and Shamir, 1961). A distributionalanalysis cannot differentiate between these proposals. The distribution is com-puted with regard to the words. The meaning of the words is not considered. Onecan observe that the utterance contains the word fischt (‘fishes’), but one cannotsee whether this word contributes the resultative semantics or not. Similarly thedistributional analysis cannot distinguish between analyses with and without anempty head. The empty head is naturally not recognizable in the signal. Theempty head is a theoretical construct and as was mentioned above an analysiswith an empty head of the kind in Figure 7 can be automatically converted intoone with a lexical rule. The argumentation for one of the latter analyses is purelytheory internal.

Concluding this subsection, we contend that Bod’s paper is a milestone in thePoverty of the Stimulus debate, but it does not and cannot show that a particularversion of constructionist theories, namely the phrasal one, is correct.

11.2 Collostructions

Stefanowitsch and Gries (2009, Section 5) assume a plugging analysis: wordsoccur in (slots provided by) a given construction if their meaning matches thatof the construction. The authors claim that their collostructional analysis hasconfirmed [the plugging analysis] from various perspectives. Stefanowitsch andGries are able to show that certain verbs occur more often than not in particularconstructions, while other verbs never occur in the respective constructions. Forinstance, give, tell, send, offer and show are attracted by the Ditransitive Con-struction, while make and do are repelled by this construction. Regarding thisdistribution the authors write:

These results are typical for collexeme analysis in that they show twothings. First, there are indeed significant associations between lexicalitems and grammatical structures. Second, these associations provideclear evidence for semantic coherence: the strongly attracted collex-emes all involve a notion of ‘transfer’, either literally or metaphori-cally, which is the meaning typically posited for the ditransitive. Thiskind of result is typical enough to warrant a general claim that col-lostructional analysis can in fact be used to identify the meaning of agrammatical construction in the first place. (Stefanowitsch and Gries,2009, p. 943)

We hope that the preceding discussion made clear that the distribution of words ina corpus cannot be seen as evidence for a phrasal analysis. The corpus study showsthat give usually is used with three arguments in a certain pattern that is typicalfor English (Subject Verb Object1 Object2) and that this verb forms a cluster with



60 12 Arguments from Neurolinguistics

other verbs that have a transfer component in their meaning. The corpus data doesnot show whether this meaning is contributed by a phrasal pattern or by lexicalentries that are used in a certain configuration.

12 Arguments from Neurolinguistics

Pulvermüller, Cappelle and Shtyrov (To appear) discuss neurolinguistic facts andrelate them to the CxG view on grammar theory. One important finding is thatdeviant words (lexical items) cause brain responses that differ in polarity frombrain responses on incorrect strings of words, that is, syntactic combinations. Thissuggests that there is indeed an empirical basis for deciding the issue.

Concerning the standard example of the caused motion construction in (93)the authors write the following:

(93) She sneezed the foam off the cappuccino.39

this constellation of brain activities may initially lead to the co-acti-vation of the verb sneeze with the DCNAs for blow and thus to thesentence mentioned. Ultimately, such co-activation of a one-placeverb and DCNAs associated with other verbs may result in the formerone-place verb being subsumed into a three-place verb category andDCNA set, a process which arguably has been accomplished for theverb laugh as used in the sequence laugh NP off the stage. (Pulver-müller, Cappelle and Shtyrov, To appear)

A DCNA is a discrete combinatorial neuronal assembly. Regarding the specificsof DCNAs the authors write that

Apart from linking categories together, typical DCNAs establish atemporal order between the category members they bind to. DCNAsthat do not impose temporal order (thus acting, in principle, as ANDunits for two constituents) are thought to join together constituentswhose sequential order is free or allow for scrambling. (Pulvermüller,Cappelle and Shtyrov, To appear)

We believe that this view is entirely compatible with the lexical view outlinedabove: the lexical item or DCNA requires certain arguments to be present. Alexical rule that relates an intransitive verb to one that can be used in the causedmotion construction is an explicit representation of what it means to activate thevalence frame of blow.

39Goldberg, 2006, p. 42



61

The authors cite earlier work (Cappelle, Shtyrov and Pulvermüller, 2010) andargue that particle verbs are lexical objects, admitting for a discontinuous realiza-tion of particle verbs despite their lexical status (p. 21). They restrict their claimto frequently occurring particle verbs. This claim is of course compatible withour assumptions here, but the differences in brain behaviour are interesting whenit comes to fully productive uses of particle verbs. For instance any mono-va-lent verb in German can be combined with the aspectual particle los: lostanzen(‘start to dance’), loslachen (‘start to laugh’), lossingen (‘start to sing’), . . . . Sim-ilarly, the combination of mono-valent verbs with the particle an with the readingdirected-towards is also productive. anfahren (‘drive towards’), anlachen (‘laughin the direction of’), ansegeln (‘sail towards’), . . . (see Stiebels, 1996 on variousproductive patterns). The interesting question is how particle verbs behave thatfollow these patterns but occur with low frequency. This is still an open questionas far as the experimental evidence is concerned, but as we argue below lexicalproposals to particle verbs as the one suggested by Müller (2003b) are compatiblewith both possible outcomes.

Summarizing the discussion so far, lexical approaches are compatible with theaccumulated neurobiological evidence and as far as particle verbs are concernedthey seem to be better suited than the phrasal proposals by Booij (2002, Section 2)and Blom (2005) (See Section 3.2 for discussion). However, in general it remainsan open question what it means to be a discontinuous lexical item. The idea of dis-continuous words is pretty old (Wells, 1947), but there have not been many formalaccounts of this idea. Nunberg, Sag and Wasow (1994) suggest a representationin a linearization-based framework of the kind that was proposed by Reape (1994)and Kathol (1995, p. 244–248) and Crysmann (2002) worked out such analyses indetail. Kathol’s lexical item for aufwachen (‘to wake up’) is given in (94):

(94)

aufwachen (following Kathol (1995, p. 246)):

. . . |HEAD 1 verb

. . . |VCOMP 〈〉

DOM

⟨

〈 wachen 〉

. . . |HEAD 1

. . . |VCOMP 〈 2 〉

⟩

©

⟨

〈 auf 〉

SYNSEM 2

. . . |HEAD

[

FLIP −

sepref

]

vc

⟩

The lexical representation contains the list-valued feature DOM that contains adescription of the main verb and the particle. The dependency between the particleand the main verb was characterized by the value of the VCOMP feature, whichis a valence feature for the selection of arguments that form a complex predicatewith their head. The shuffle operator © concatenates two lists without specifyingan order between the elements of the two lists, that is, both wachen, auf and auf,



62 12 Arguments from Neurolinguistics

wachen are possible. The little marking vc is an assignment to a topological fieldin the clause.

Müller (2007a) criticized such linearization-based proposals since it is unclearhow analyses that claim that the particle is just linearized in the domain of itsverb can account for sentence like (95), in which complex syntactic structuresare involved. German is a V2 language and the fronting of a constituent intothe position before the finite verb is usually described as some sort of nonlocaldependency, that is, even authors who assume linearization-based analyses do notassume that the initial position is filled by simple reordering of material (Kathol,2001; Müller, 1999, 2002a; Bjerre, 2006).40

(95) a. [vf [mf Denthe

Atem]breath

[vc an]]PART

hieltheld

diethe

ganzewhole

Judenheit.41

Jewish.community‘The whole Jewish community held their breath.’

b. [vf [mf Wieder]again

[vc an]]PART

tretenkick

auchalso

diethe

beidentwo

Sozialdemokraten.42

social.democrats‘The two Social Democrates are also running for office again.’

40Kathol (1995, Section 6.3) suggested such an analysis, but later changed his view. Wetta(2011) also assumes a purely linearization-based approach. He assumes that sentences in whichmultiple constituents are fronted (Müller, 2003b) are analyzed in such a way that more than onelinearization object are inserted as one single object into the position before the finite verb. Thisfails to account for multiple frontings that cross a clause boundary as in the examples in (i) dis-cussed by Fanselow (1993, p. 67):

(i) a. Derthe

MariaMaria

dasthe

Buchbook

wennif

duyou

denkstthink

daßthat

duyou

gebengive

darfstbe.allowed.to

bistare

duyou

schönpretty

blöd.stupid‘You are pretty stupid if you think you are allowed to give Maria the book.’

b. Derthe

MariaMaria

einena

Ringring

glaubbelieve

ichI

nichtnot

daßthat

erhe

jeever

schenkengive.as.apresent

wird.will

‘I do not believe that he will ever give Maria a ring as a present.’

If such sentences are to be analyzed as verb second sentences involving a dislocation mechanism– as also assumed by Wetta for non-local extraction (p. 265), there has to be a connection betweena single element in the embedded clause and the fronted constituents der Maria das Buch andder Maria einen Ring. However, no such single projection exists in linearization-based proposals,since der Maria is the dative object and das Buch is the accusative object and the two do not forma constituent on the level of structure that is relevant for extraction. In earlier work Müller (2000)suggested two nonlocal dependencies for the analysis of multiple frontings, but this was revisedlater and superseded by an analysis that integrates nicely into the rest of German grammar (Müller,2005a,b).

41Lion Feuchtwanger, Jud Süß, p. 276, quoted from Grubacic, 1965, p. 56.42taz, bremen, 24.05.2004, p. 21.



63

c. [vf [vc Los]PART

[nf damit]]there.with

gehtwent

esit

schonalready

amat.the

15.15

April.43

April‘It already started on April the 15th.’

The conclusion that has to be drawn from examples like (95) is that particles inter-act in complex ways with the syntax of sentences. This is captured by the lexicaltreatment that was suggested in Müller, 2002a, Chapter 6 and Müller, 2003b: Themain verb selects for the verbal particle. By assuming that wachen selects forauf the tight connection between verb and particle is represented.44 Such a lexi-cal analysis provides an easy way to account for fully intransparent particle verbslike an-fangen (‘to begin’). However, we also argued for a lexical treatment oftransparent particle verbs like losfahren (‘to start to drive’) and jemanden/etwasanfahren (‘drive directed towards somebody/something’). The analysis involvesa lexical rule that licences a verbal item that selects for an adjunct particle. Theparticles an and los can modify verbs and contribute arguments (in the case of an)and their semantics. This analysis can be shown to be compatible with the neuro-mechanical findings: if it is the case that even transparent particle verb combina-tions with low frequency are stored than the rather general lexical rule that wassuggested by Müller is the generalization of the relation between a large amount oflexical particle verb items and their respective main verb. The individual particleverbs would be special instantiations that have the form of the particle specifiedas it is also the case for non-transparent particle verbs like anfangen. If it shouldturn out that productive particle verb combinations with particle verbs of low fre-quency cause syntactic reflexes in the brain, this could be explained as well: Thelexical rule licences an item that selects for an adverbial element. This selectionwould then be seen as parallel to the relation between the determiner and the nounin the NP der Mut (‘the courage’), which (Cappelle et al., 2010, p. 191) discussas an example of a syntactic combination. Note that Müller’s analysis is alsocompatible with another observation made by Shtyrov, Pihko and Pulvermüller(2005): Morphological affixes also cause the lexical reflexes. In Müller’s analy-sis the stem of the main verb is related to another stem that selects for a particle.This stem can be combined with (derivational and inflectional) morphological af-fixes causing the lexical activation pattern in the brain. After this combination theverb is combined with the particle and the dependency can be either a lexical or asyntactic one, depending on the results of the experiments to be carried out. The

43taz, 01.03.2002, p. 8.44Cappelle et al. (2010, p. 197) write: the results provide neurophysiological evidence that

phrasal verbs are lexical items. Indeed, the increased activation that we found for existing phrasalverbs, as compared to infelicitous combinations, suggests that a verb and its particle together formone single lexical representation, i. e. a single lexeme, and that a unified cortical memory circuitexists for it, similar to that encoding a single word We believe that Müller’s analysis is compatiblewith this statement.



64 13 Conclusion

analysis is compatible with both results.Note that Müller’s analysis allows the principle of lexical integrity to be main-

tained. We therefore do not follow (Cappelle et al., 2010, p. 198), who claim thatthey provide proof that potentially separable multi-word items can nonethelessbe word-like themselves, and thus against the validity of a once well-establishedlinguistic principle, the Lexical Integrity Principle. We agree that non-transparentparticle verbs are multi-word lexemes, but the existence of multi-word lexemesdoes not show that syntax has access to the word internal morphological struc-ture. The parallel between particle verbs and clearly phrasal idioms was discussedin Müller, 2002a,b and it was concluded that idiom-status is irrelevant for thequestion of wordhood. Since the interaction of clearly phrasal idioms with deriva-tional morphology as evidenced by examples like (96) did not force grammariansto give up on Lexical Integrity, it can be argued that particle verbs are not con-vincing evidence for giving up the Lexical Integrity Principle either.45

(96) a. Erhe

hathas

insin.the

Grasgras

gebissen.bit

‘He bit the dust.’

b. „HeathHeath

Ledger“Ledger

kanncan

ichI

nichtnot

einmaleven

schreiben,write

ohnewithout

dassthat

mirme

seinhis

insin.the

Gras-Gebeißegrass.biting

wiederagain

soso

wahnsinnigcrazy

leidsorrow

tut46

does‘I cannot even write “Heath Ledger” without being sad again about hisrepeated biting the dust.’

The example in (96b) involves the discontinuous derivation with the circumfix Ge--e (Lüdeling, 2001, Section 3.4.3; Müller, 2002a, p. 324–327, 372–377; Müller,2003b, Section 2.2.1, Section 5.2.1). Still the parts of the idiom ins Gras beiß- arepresent and with them the idiomatic reading. See Sag, 2007 for a lexical analysisof idioms that can explain examples like (96).

13 Conclusion

We have shown in this paper that there are no compelling arguments for assum-ing phrasal argument structure constructions, but that there are several argumentsagainst them. Assuming a lexical or — in the terminology of Goldberg (To appear)— template-based view solves all the problems that arise for phrasal approaches.

Furthermore we showed that radically underspecified approaches in the senseof Borer (2005), Haugereid (2007), and Lohndal (2012) are not restricted enough.

45However, see Booij, 2009 for some challenges to Lexical Integrity.46http://www.coffee2watch.at/egala. 23.03.2012


http://www.coffee2watch.at/egala


65

The only way to establish the necessary restrictions is a lexical representation,since the information that has to be captured is in part lexeme dependent.

14 Appendix: subject idioms

A very frequent claim throughout the literature is that subjects have a specialstatus, that subjects are not selected by their head (Chomsky, 1981, p. 26–28;Hoekstra, 1987, p. 33). However, Bresnan (1982a, Chapter 2) already showed thatmany of the claims that were made to support the non-selection of the subject areempirically wrong.

The claim that the subject is not selected is supported by the claim that thereare no idioms with a subject as part of the idiom and an object slot that can be filledfreely (Marantz, 1981, p. 50–51). Nunberg et al. (1994, p. 526) cited example 97a,among others, and Bresnan (1982a, p. 349–350) cited 97b and 97c as examples inwhich the subject is a fixed part of the idiom and there are open slots for non-subjects:

(97) a. A little bird told X that S.‘X heard the rumor that S’)

b. The cat’s got x’s tongue.‘X cannot speak.’

c. What’s eating x?‘Why is X so galled?’

Marantz (1984, p. 29) remarks that examples like (97b) are irrelevant for the dis-cussion of his claims, since the free element is not the object. Regarding (97c),Marantz writes:

From the point of view of the present theory, it is important that thisapparent subject idiom has no S-internal syntax, for it is preciselyS-internal syntax that is at issue. What’s eating NP? is not a com-bination of subject and verb, forming a predicate on the object, butrather a combination of wh-question syntax, progressive aspect, plussubject and verb—that is, a complete sentence frame—with an openslot for an argument. (Marantz, 1984, p. 27)

Interestingly this points towards a Construction Grammar analysis that assumesthat even complex linguistic objects can be stored. However, such approachesassume that complex linguistic objects do have internal structure. If one ignoredthis internal structure, it would not be possible to explain in what respect (97c) issimilar to other English sentences and why (97c) is structured internally accordingto normal syntactic laws of English. Therefore, idiomatic phrases do get assigned



66 14 Appendix: subject idioms

an internal structure in frameworks like TAG (Abeillé and Schabes, 1989), HPSG(Krenn and Erbach, 1994; Sailer, 2000; Soehn and Sailer, 2008; Soehn, 2006; Sag,2007; Richter and Sailer, 2009), and Simpler Syntax (Culicover and Jackendoff,2005).

Marantz explains some of the examples that were provided by Bresnan awayby assuming that the subjects are not subjects underlyingly but objects since theverbs in the idioms are unaccusative verbs.

The claim that subjects are never part of idioms that have an open slot iswidespread (see for instance den Dikken, 1995, p. 92, Kratzer, 1996, p. 112–116)and is also explicitly made for German for instance by Grewendorf (2002, p. 50).That the claim is not correct for German and hence cannot hold universally isshown by examples like (98) and (99), some of which were already discussed byReis (1982, p. 178):47

(98) a. weilbecause

ihnhim.ACC

derthe

Haferoats.NOM

stichtpricks

‘because he is feeling his oats’

b. weilbecause

ihnhim.ACC

derthe

Teufeldevil.NOM

reitetrides

‘because he had the devil in him’

c. weilbecause

ihmhim.DAT

alleall

Fellefurs.NOM

davonschwimmenaway.swim

‘because all his hopes were dashed’

d. weilbecause

ihnhim.ACC

derthe

Schlagstroke.NOM

trifftstrikes

‘because he was flabbergasted’

e. weilbecause

ihnhim.ACC

diethe

Wutrage.NOM

packtgrabs

‘because he flies into a rage’

f. weilbecause

ihmhim.DAT

derthe

Kopfhead.NOM

rauchtsmokes

‘because his head is spinning’

g. weilbecause

keinno

Hahncock

danachthere.after

krähtcrows

‘because no one cares two hoots about that’

h. weilbecause

nachafter

ihmhim

keinno

Hahncock.nom

krähtcrows

“because no one cares two hoots about him’47(98a), (98b), (98c), (99a) und (99b) are due to Reis (1982, p. 178). See also Haider (1993,

p. 173) for examples that are similar to those in (98).



67

i. weilbecause

ihnhim.ACC

derthe

Eseldonkey.NOM

imin.the

Galoppgallop

verlorenlost

hathas

‘He is out of the question.’ (according to Duden Redewendungen)

j. weilbecause

diethe

Spatzensparrows.NOM

dasthat.ACC

vomfrom.the

Dachroof

gepfiffenwhistled

habenhave

‘because it was all over town’

(99) a. weilbecause

ihmhim.DAT

derthe

Geduldsfadenpatience.twist.NOM

reißttears

‘because his patience gave out’

b. weilbecause

ihmhim.DAT

derthe

Kragencollar

geplatztis

ist

‘because he blew his top’

c. weilbecause

ihmhim

eina

Steinstone

vomfrom.the

Herzenheart

fälltfalls

‘because that’s a load off his mind’

d. weilbecause

beiat

ihmhim

Hopfenhop

undand

Malzmalt

verlorenlost

istis

‘because he is a hopeless case’

Marantz (1997, p. 208–209) formulates a variant of his earlier claim stating thatan agent may not be a fixed part of an idiom. The examples in (98b) and (98g)show that even this more restricted version of the no-subject in idioms claim iswrong.

Scherpenisse (1986, p. 89) working on German repeats Marantz’ claim forEnglish (1984) that idioms that contain subjects as fixed part can only be formedwith unaccusative verbs, that is, the subjects are claimed to be underlying objects.If this claim were true there should not be passivizable idioms with fixed sub-jects, since the passive is the suppression of the subject. If there is no underlyingsubject, passive cannot apply. However, if one looks at the above examples onnotes that those idioms in (98) that contain verbs that can be passivized in theirliteral meaning can be passivized in the idiom reading as well. It is well-knownfrom other idioms that passivization is sometimes marked. Many of the examplescontain inanimate subjects which also influences passivizability.48 Examples ofpassivizations of the idiomatic variants are shown in (100) and (102a, b):

(100) a. Was schon von den Dächern gepfiffen wurde, jetzt ist es amtlich:[. . . ]49

48But see Müller, 2002a, Chapter 3.1.2 for passives with inanimate subjects.49http://witch.muensterland.org/2003/04/18.html. 05.01.2007.


http://witch.muensterland.org/2003/04/18.html


68 14 Appendix: subject idioms

b. Was vor wenigen Wochen nur vereinzelt von den Dächern gepfiffenwurde, nahm in der letzten Woche konkrete Formen an: [. . . ]50

c. In einer Pressemitteilung bestätigte die WWF am 23.03.2001 offizielldie Nachricht, die von allen Spatzen längst von den Dächern gepfiffenwurde.51

Example (101) shows a state passive of (98a), and the examples in (102) are pas-sives of (98e), (102c) containing a prenominal participle with passive argumentstructure:

(101) Vom Hafer gestochen52

(102) a. Und doch kann sich kaum eine Frau davon freisprechen, daß sie vonEmpörung oder gar heiliger Wut gepackt wird, sobald sie eines Mannesauf ihrem Territorium ansichtig wird.53

b. [. . . ] Michaela Gerg, die nach einem verkorksten ersten Lauf von Wutgepackt wurde, im zweiten Durchgang Bestzeit fuhr und noch auf densiebten Rang kam.54

c. Mit 2:6 gab sie den ersten Satz übernervös ab, drehte daraufhin vonWut gepackt teuflisch auf und holte sich den zweiten Satz unter Zuhil-fenahme göttlicher Passierschläge und perfekter Lobs ebenfalls mit 6:2.55

(103) and (104) show further passivizations and prenominal participles with pas-sive argument structure:

(103) a. „Iordannis wird vom Teufel geritten“, witzelte Vangelis.56

b. Wird Gerda vom Teufel geritten?57

c. Er wurde verhext, oder sie wurde vom Teufel geritten.58

(104) a. viele vom Schlag getroffene und Lahme aber wurden geheilt.59

b. Jeffrey, der Sohn des vom Schlag Getroffenen, besucht seinen Vater imKrankenhaus.60

50http://heim.at/paysuspends/kulturmedien/saban.htm.51http://people.freenet.de/wwf-hp/WWFschlucktWCW.htm. 05.01.2007.52taz, 06.04.2000, S. 20.53taz bremen, 29.11.1997, S. 26.54taz, 29.01.1990, S. 13.55taz, 22.10.1991, S. 13.56Jentzsch, Kerstin, Seit die Götter ratlos sind, München: Heyne 1999 [1994] S. 227.57Dietlof Reiche, Unterwegs mit Gerda, in: Die Zeit 17.09.1998, S. 71.58Schwanitz, Dietrich, Männer, Frankfurt a.M.: Eichborn 2001, S. 241.59http://members.tirol.com/vineyard.grk/predigt.html. 22.09.2003.60Frankfurter Allgemeine Zeitung; 13.11.1986. http://www.davidlynch.de/bluefaz86.html.

22.09.2003.


http://heim.at/paysuspends/kulturmedien/saban.htm

http://people.freenet.de/wwf-hp/WWFschlucktWCW.htm

http://www.davidlynch.de/bluefaz86.html


REFERENCES 69

This shows that idioms like Die Spatzen pfeifen X von den Dächern, X packt dieWut, der Teufel reitet X and Y trifft der Schlag are syntactically active. A theoryabout human language should capture that the structures for active sentences arerelated to the ones for passive sentences in the same way as this is captured fornon-idiomatic sentences.

Sternefeld (1985, p. 435) notes that the nominatives in Marga Reis’ idiom ex-amples behave like objects since they are adjacent to the verb and therefore havethe same ordering properties as objects. Therefore Sternefeld assumes an incor-poration analysis for such idioms, that is, he assumes that the verb and the nomi-native form a fixed unit. The examples in (98g) and (98j) show that idiom parts donot have to be adjacent to the verb. Apart from this, the idiom can be passivized,which shows that it would be inadequate to list the idiom as a fixed unchangableunit in the lexicon. die Spatzen (‘the sparrows’) is a syntactically normal subjectthat behaves normally in transformations like passivization.

We can conclude this discussion as follows:

• There exist idioms with the subject as part of the idiom and an open slot forthe object. (accusative object, dative object/possessive dative, prepositionalobject, and probably other grammatical functions as well)

• Some of these idioms contain verbs that are not unaccusative.

• There are idioms with fixed subject that can be passivized or form adjectivalpassive participles.

• There are idioms with fixed subject whose parts (including the subject) canbe reordered.

This shows that these idioms are not fixed entities with one open slot but ratherunits that have an active syntactic live. Hence they contradict Marantz’ claim inthe quote above. Since this claim is falsified for one language it cannot hold forall languages. Since approaches to idioms that establish the idiomaticity of anexpression after the syntactic analysis were shown to be inadequate (Keil, 1997),the discussion suggests that there has to be a selection relation between heads andtheir subjects in order to capture the idioms discussed above.

References

Abeillé, Anne and Schabes, Yves. 1989. Parsing Idioms in Lexicalized TAG. InHarold Somers and Mary McGee Wood (eds.), Proceedings of the Fourth Con-ference of the European Chapter of the Association for Computational Linguis-tics, pages 1–9, Manchester, England: Association for Computational Linguis-tics.



70 REFERENCES

Ajdukiewicz, Kasimir. 1935. Die syntaktische Konnexität. Studia Philosophica 1,1–27.

Alsina, Alex. 1996. Resultatives: A Joint Operation of Semantic and SyntacticStructures. In Miriam Butt and Tracy Holloway King (eds.), Proceedings of theLFG96 Conference, Rank Xerox, Grenoble, Stanford, CA: CSLI Publications.http://cslipublications.stanford.edu/LFG/1/lfg1.html, 06.07.2002.

Asudeh, Ash, Dalrymple, Mary and Toivonen, Ida. 2008. Constructions withLexical Integrity: Templates as the Lexicon-Syntax Interface. In Miriam Buttand Tracy Holloway King (eds.), Proceedings of the LFG 2008 Conference,Stanford, CA: CSLI Publications. http://cslipublications.stanford.edu/LFG/13/,27.09.2012.

Bach, Emmon. 1976. An Extension of Classical Transformation Grammar. InProblems in Linguistic Metatheory, Proceedings of the 1976 Conference at theMichigan State University, pages 183–224.

Bar-Hillel, Yehoshua, Perles, Micha A. and Shamir, Eliahu. 1961. On For-mal Properties of Simple Phrase-Structure Grammars. Zeitschrift für Phonetik,Sprachwissenschaft und Kommunikationsforschung 14(2), 143–172.

Bergen, Benjamin K. and Chang, Nancy. 2005. Embodied Construction Grammarin Simulation-Based Language Understanding. In Jan-Ola Östman and Mir-jam Fried (eds.), Construction Grammars: Cognitive Grounding and Theoreti-cal Extensions, pages 147–190, Amsterdam and Philadelphia: John BenjaminsPublishing Co.

Bjerre, Tavs. 2006. Object Positions in a Topological Sentence Model for Dan-ish – a Linearization-Based HPSG Approach. Presentation at Ph.D.-courseat Sandbjerg, Denmark. http://www.hum.au.dk/engelsk/engsv/objectpositions/workshop/Bjerre.pdf, 23.12.2011.

Blom, Corrien. 2005. Complex Predicates in Dutch. Synchrony and Di-achrony. LOT Dissertation Series, No. 111, Utrecht University. http://www.lotpublications.nl/publish/issues/BlomC/, 31.08.2005.

Bod, Rens. 2009a. Constructions at Work or at Rest? Cognitive Linguistics 20(1),129–134.

Bod, Rens. 2009b. From Exemplar to Grammar: Integrating Analogy and Prob-ability in Language Learning. Cognitive Science 33(4), 752–793. http://staff.science.uva.nl/~rens/analogy.pdf, 15.07.2008.

Booij, Geert. To appear. Construction Morphology. In Andrew Spencer andArnold Zwicky (eds.), The Handbook of Morphology, London: BlackwellPublishing Ltd, second edition. http://www.hum2.leidenuniv.nl/booijge/pdf/Spencer%20Zwicky%20handbook%20Construction%20morphology.pdf,30.05.2012.

Booij, Geert E. 2002. Separable Complex Verbs in Dutch: A Case of PeriphrasticWord Formation. In Dehé et al. (2002), pages 21–41.


http://cslipublications.stanford.edu/LFG/1/lfg1.html

http://cslipublications.stanford.edu/LFG/13/

http://www.hum.au.dk/engelsk/engsv/objectpositions/workshop/Bjerre.pdf

http://www.hum.au.dk/engelsk/engsv/objectpositions/workshop/Bjerre.pdf

http://www.lotpublications.nl/publish/issues/BlomC/

http://www.lotpublications.nl/publish/issues/BlomC/

http://staff.science.uva.nl/~rens/analogy.pdf

http://staff.science.uva.nl/~rens/analogy.pdf

http://www.hum2.leidenuniv.nl/booijge/pdf/Spencer%20Zwicky%20handbook%20Construction%20morphology.pdf

http://www.hum2.leidenuniv.nl/booijge/pdf/Spencer%20Zwicky%20handbook%20Construction%20morphology.pdf


REFERENCES 71

Booij, Geert E. 2009. Lexical Integrity as a Formal Universal: A ConstructionistView. In Sergio Scalise, Elisabetta Magni and Antonietta Bisetto (eds.), Univer-sals of Language Today, Studies in Natural Language and Linguistic Theory,No. 76, pages 83–100, Springer Verlag.

Borer, Hagit. 1994. The Projection of Arguments. In E. Benedicto and J. Run-ner (eds.), Functional Projections, UMass Occasional Papers in Linguistics(UMOP), No. 17, pages 19–47, Massachusetts: University of MassachusettsGraduate Linguistic Student Association.

Borer, Hagit. 2003. Exo-Skeletal vs. Endo-Skeletal Explanations: Syntactic Pro-jections and the Lexicon. In John Moore and Maria Polinsky (eds.), The Natureof Explanation in Linguistic Theory, pages 31–67, Stanford, CA: CSLI Publi-cations.

Borer, Hagit. 2005. Structuring Sense, volume I. In Name Only. Oxford: OxfordUniversity Press.

Bouma, Gosse, Malouf, Robert and Sag, Ivan A. 2001. Satisfying Constraintson Extraction and Adjunction. Natural Language and Linguistic Theory 19(1),1–65. http://ftp-linguistics.stanford.edu/sag/bms-nllt.pdf, 31.05.2010.

Bresnan, Joan. 1976. Nonarguments for Raising. Linguistic Inquiry 7(3), 485–502.

Bresnan, Joan. 1982a. Control and Complementation. Linguistic Inquiry 13(3),343–434.

Bresnan, Joan (ed.). 1982b. The Mental Representation of Grammatical Relations.MIT Press Series on Cognitive Theory and Mental Representation, Cambridge,MA/London: MIT Press.

Bresnan, Joan. 1982c. The Passive in Lexical Theory. In Bresnan (1982b), pages3–86.

Bresnan, Joan. 1994. Locative Inversion and the Architecture of Universal Gram-mar. Language 70(1), 72–131.

Bresnan, Joan. 2001. Lexical-Functional Syntax. Oxford, UK/Cambridge, USA:Blackwell.

Bresnan, Joan and Kaplan, Ronald M. 1982. Introduction: Grammars as MentalRepresentations of Language. In Bresnan (1982b), pages xvii–lii.

Bresnan, Joan and Mchombo, Sam A. 1995. The Lexical Integrity Principle: Evi-dence from Bantu. Natural Language and Linguistic Theory 13, 181–254.

Briscoe, Ted J. and Copestake, Ann. 1999. Lexical Rules in Constraint-BasedGrammar. Computational Linguistics 25(4), 487–526. http://acl.ldc.upenn.edu/J/J99/J99-4002.pdf, 06.04.2004.

Calcagno, Mike. 1995. Interpreting Lexical Rules. In Glyn V. Morrill andRichard T. Oehrle (eds.), Proceedings of the Formal Grammar Conference,Barcelona, Spain.

Calcagno, Mike and Pollard, Carl J. 1995. Lexical Rules in HPSG: What are they?,


http://ftp-linguistics.stanford.edu/sag/bms-nllt.pdf

http://acl.ldc.upenn.edu/J/J99/J99-4002.pdf

http://acl.ldc.upenn.edu/J/J99/J99-4002.pdf


72 REFERENCES

Department of Linguistics, Ohio State University, Columbus, ms.Candito, Marie-Hélène. 1996. A Principle-Based Hierarchical Representation of

LTAGs. In Tsuji (1996), pages 194–199.Cappelle, Bert, Shtyrov, Yury and Pulvermüller, Friedemann. 2010. Heating up

or cooling up the Brain? MEG Evidence that Phrasal Verbs are Lexical Units.Brain and Language 115, 189–201.

Carlson, Gregory N. and Tanenhaus, Michael K. 1988. Thematic Roles and Lan-guage Comprehension. In Wendy Wilkins (ed.), Thematic Relations, Syntaxand Semantics, No. 21, pages 263–289, San Diego: Academic Press.

Chomsky, Noam. 1957. Syntactic Structures. Janua Linguarum / Series Minor,No. 4, The Hague/Paris: Mouton.

Chomsky, Noam. 1965. Aspects of the Theory of Syntax. Cambridge, MA: MITPress.

Chomsky, Noam. 1970. Remarks on Nominalization. In Roderick A. Jacobs andPeter S. Rosenbaum (eds.), Readings in English Transformational Grammar,Chapter 12, pages 184–221, Waltham, MA/Toronto/London: Ginn and Com-pany.

Chomsky, Noam. 1971. Problems of Knowledge and Freedom. London: Fontana.Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris

Publications.Clément, Lionel and Kinyon, Alexandra. 2003. Generating Parallel Multilingual

LFG-TAG Grammars from a MetaGrammar. In Erhard Hinrichs and Dan Roth(eds.), Proceedings of the 41st Annual Meeting of the Association for Compu-tational Linguistics, pages 184–191, Sapporo, Japan: Association for Compu-tational Linguistics.

Copestake, Ann. 1992. The Representation of Lexical Semantic Information. Cog-nitive Science Research Papers 280, University of Sussex.

Copestake, Ann and Briscoe, Ted J. 1992. Lexical Operations in a UnificationBased Framework. In James Pustejovsky and Sabine Bergler (eds.), LexicalSemantics and Knowledge Representation, Lecture Notes in Artificial Intelli-gence, No. 627, pages 101–119, Berlin: Springer Verlag. http://www.cl.cam.ac.uk/Research/NL/acquilex/papers.html, 18.08.2002.

Croft, William. 2001. Radical Construction Grammar. Syntactic Theory in Typo-logical Perspective. Oxford University Press.

Croft, William. 2003. Lexical Rules vs. Constructions: A False Dichotomy. InHubert Cuyckens, Thomas Berg, René Dirven and Klaus-Uwe Panther (eds.),Motivation in Language: Studies in honour of Günter Radden, pages 49–68,Amsterdam: John Benjamins Publishing Co.

Crysmann, Berthold. 2002. Constraint-Based Coanalysis: Portuguese Cliticisa-tion and Morphology-SyntaxInteraction in HPSG. Saarbrücken Dissertationsin Computational Linguisticsand Language Technology, No. 15, Saarbrücken:


http://www.cl.cam.ac.uk/Research/NL/acquilex/papers.html

http://www.cl.cam.ac.uk/Research/NL/acquilex/papers.html


REFERENCES 73

Deutsches Forschungszentrum für Künstliche Intelligenz undUniversität desSaarlandes.

Culicover, Peter W. and Jackendoff, Ray S. 2005. Simpler Syntax. Oxford: OxfordUniversity Press.

Dabrowska, Ewa. 2001. From Formula to Schema: The Acquisition of EnglishQuestions. Cognitive Linguistics 11(1–2), 83–102.

Davidson, Donald. 1967. The Logical Form of Action Sentences. In NicholasRescher (ed.), The Logic of Decision and Action, pages 81–95, Pittsburg Uni-versity Press, zitiert von Rothstein85a:111-112, Kaufmann95a:14, nicht CoLi,DFKI, SULB.

De Kuthy, Kordula. 2002. Discontinuous NPs in German. Studies in Constraint-Based Lexicalism, No. 14, Stanford, CA: CSLI Publications.

De Kuthy, Kordula and Meurers, Walt Detmar. 2001. On Partial ConstituentFronting in German. Journal of Comparative Germanic Linguistics 3(3),143–205. http://www.ling.ohio-state.edu/~kdk/papers/dekuthy-meurers-jcgs.html, 19.08.2002.

Dehé, Nicole, Jackendoff, Ray S., McIntyre, Andrew and Urban, Silke (eds.).2002. Verb-Particle Explorations. Interface Explorations, No. 1, Berlin/NewYork, NY: Mouton de Gruyter.

den Dikken, Marcel. 1995. Particles. On the Syntax of Verb-Particle, Triadic, andCausative Constructions. Oxford Studies in Comparative Syntax, New York,Oxford: Oxford University Press.

Diderichsen, Poul. 1957. Elementær dansk grammatik. København: Gyldendal.Dowty, David R. 1978. Governed Transformations as Lexical Rules in a Montague

Grammar. Linguistic Inquiry 9(3), 393–426.Dowty, David R. 1989. On the Semantic Content of the Notion ‘Thematic Role’.

In G. Chierchia, B. Partee and R. Turner (eds.), Properties, Types and Mean-ing, volume 2, pages 69–130, Dordrecht/Boston/London: Kluwer AcademicPublishers.

Embick, David. 2004. On the Structure of Resultative Participles in English. Lin-guistic Inquiry 35(3), 355–392.

Fanselow, Gisbert. 1993. Die Rückkehr der Basisgenerierer. Groninger Arbeitenzur Germanistischen Linguistik 36, 1–74.

Fanselow, Gisbert. 2001. Features, θ-Roles, and Free Constituent Order. Linguis-tic Inquiry 32(3), 405–437.

Fanselow, Gisbert. 2002. Against Remnant VP-Movement. In Artemis Alexiadou,Elena Anagnostopoulou, Sjef Barbiers and Hans-MartinGärtner (eds.), Dimen-sions of Movement. From Features to Remnants, Linguistik Aktuell/LinguisticsToday, No. 48, pages 91–127, Amsterdam and Philadelphia: John BenjaminsPublishing Co.

Fillmore, Charles J. 1999. Inversion and Constructional Inheritance. In Gert We-


http://www.ling.ohio-state.edu/~kdk/papers/dekuthy-meurers-jcgs.html

http://www.ling.ohio-state.edu/~kdk/papers/dekuthy-meurers-jcgs.html


74 REFERENCES

belhuth, Jean-Pierre Koenig and Andreas Kathol (eds.), Lexical and Construc-tional Aspects of Linguistic Explanation, Studies in Constraint-Based Lexical-ism, No. 1, pages 113–128, Stanford, CA: CSLI Publications.

Fillmore, Charles J., Kay, Paul and O’Connor, Mary Catherine. 1988. Regularityand Idiomaticity in Grammatical Constructions: The Case of LetAlone. Lan-guage 64(3), 501–538.

Gazdar, Gerald. 1982. Phrase Structure Grammar. In Pauline Jacobson and Ge-offrey K. Pullum (eds.), The Nature of Syntactic Representation, Synthes Lan-guage Library, No. 15, Dordrecht/Boston/London: D. Reidel Publishing Com-pany.

Gazdar, Gerald, Klein, Ewan, Pullum, Geoffrey K. and Sag, Ivan A. 1985. Gener-alized Phrase Structure Grammar. Cambridge, MA: Harvard University Press.

Geach, Peter Thomas. 1970. A Program for Syntax. Synthese 22, 3–17.Goldberg, Adele E. 1995. Constructions. A Construction Grammar Approach to

Argument Structure. Cognitive Theory of Language and Culture, Chicago/Lon-don: The University of Chicago Press.

Goldberg, Adele E. 2003. Words by Default: the Persian Complex Predicate Con-struction. In Elaine J. Francis and Laura A. Michaelis (eds.), Mismatch: Form-function Incongruity and the Architecture of Grammar, CSLI Lecture Notes,No. 163, pages 117–146, Stanford, CA: CSLI Publications.

Goldberg, Adele E. 2006. Constructions at Work. The Nature of Generalization inLanguage. Oxford Linguistics, Oxford, New York: Oxford University Press.

Goldberg, Adele E. To appear. Argument Structure Constructions vs. LexicalRules or Derivational Verb Templates. Mind and Language .

Goldberg, Adele E., Casenhiser, Devin and Sethuraman, Nitya. 2004. LearningArgument Structure Generalizations. Cognitive Linguistics 15(3), 289–316.

Goldberg, Adele E. and Jackendoff, Ray S. 2004. The English Resultative as aFamily of Constructions. Language 80(3), 532–568.

Grewendorf, Günther. 1993. German. A Grammatical Sketch. In Joachim Jacobs,Arnim von Stechow, Wolfgang Sternefeld and Theo Vennemann (eds.), Syn-tax – Ein internationales Handbuch zeitgenössischer Forschung, volume 9.2 ofHandbücher zur Sprach- und Kommunikationswissenschaft, pages 1288–1319,Berlin: Walter de Gruyter Verlag.

Grewendorf, Günther. 2002. Minimalistische Syntax. UTB für Wissenschaft: Uni-Taschenbücher, No. 2313, Tübingen, Basel: A. Francke Verlag GmbH.

Grubacic, Emilija. 1965. Untersuchungen zur Frage der Wortstellung in derdeutschen Prosadichtung derletzten Jahrzehnte. Ph. D.thesis, PhilosophischeFakultät, Zagreb.

Gunji, Takao. 1986. Subcategorization and Word Order. In William J. Poser (ed.),Papers from the Second International Workshop on Japanese Syntax, pages 1–21, Stanford, CA: CSLI Publications.



REFERENCES 75

Haider, Hubert. 1985. The Case of German. In Jindrich Toman (ed.), Studies inGerman Grammar, Studies in Generative Grammar, No. 21, pages 23–64, Dor-drecht: Holland, Cinnaminson: U.S.A.: Foris Publications.

Haider, Hubert. 1990. Topicalization and other Puzzles of German Syntax. InGünther Grewendorf and Wolfgang Sternefeld (eds.), Scrambling and Barri-ers, Linguistik Aktuell/Linguistics Today, No. 5, pages 93–112, Amsterdamand Philadelphia: John Benjamins Publishing Co.

Haider, Hubert. 1993. Deutsche Syntax – generativ. Vorstudien zur Theorie einerprojektiven Grammatik. Tübinger Beiträge zur Linguistik, No. 325, Tübingen:Gunter Narr Verlag.

Hale, Kenneth and Keyser, Samuel Jay. 1993. On Argument Structure and the Lex-ical Expression of Syntactic Relations. In Kenneth Hale and Samuel Jay Keyser(eds.), The View from Building 20: Essays in Linguistics in Honor of SylvainBromberger, Current Studies in Linguistics, No. 24, pages 53–109, Cambridge,MA/London: MIT Press.

Hale, Kenneth and Keyser, Samuel Jay. 1997. On the Complex Nature of Sim-ple Predicators. In Alex Alsina, Joan Bresnan and Peter Sells (eds.), ComplexPredicates, CSLI Lecture Notes, No. 64, pages 29–65, Stanford, CA: CSLIPublications.

Harley, Heidi and Noyer, Rolf. 2000. Formal versus encyclopedic properties ofvocabulary: Evidence from nominalizations. In Peeters (2000), pages 349–374.

Haugereid, Petter. 2007. Decomposed Phrasal Constructions. In Müller (2007b).http://cslipublications.stanford.edu/HPSG/8/, 29.07.2007.

Haugereid, Petter. 2009. Phrasal Subconstructions: A Constructionalist Gram-mar Design, Exemplified with Norwegian and English. Ph. D.thesis, NorwegianUniversity of Science and Technology.

Heinz, Wolfgang and Matiasek, Johannes. 1994. Argument Structure and CaseAssignment in German. In Nerbonne et al. (1994), pages 199–236.

Hinrichs, Erhard W. and Nakazawa, Tsuneko. 1994. Partial-VP and Split-NP Top-icalization in German—An HPSG Analysis. In Erhard W. Hinrichs, Walt Det-mar Meurers and Tsuneko Nakazawa (eds.), Partial-VP and Split-NP Topical-ization in German—An HPSG Analysis and itsImplementation, Arbeitspapieredes SFB 340, No. No. 58, Eberhard-Karls-Universität Tübingen.

Hoeksema, Jack. 1991. Theoretische Aspekten van Partikelvooropplaatsing.TABU Bulletin voor Taalwetenschap 21(1), 18–26.

Hoekstra, Teun. 1987. Transitivity. Grammatical Relations in Government-Bind-ing Theory. Dordrecht: Holland, Cinnaminson: U.S.A.: Foris Publications.

Hoffman, Beryl Ann. 1995. The Computational Analysis of the Syntax and Inter-pretation of “Free” Word Order in Turkish. Ph. D.thesis, University of Pennsyl-vania.

Höhle, Tilman N. 1978. Lexikalische Syntax: Die Aktiv-Passiv-Relation und an-


http://cslipublications.stanford.edu/HPSG/8/


76 REFERENCES

dere Infinitkonstruktionen imDeutschen. Linguistische Arbeiten, No. 67, Tübin-gen: Max Niemeyer Verlag.

Holmberg, Anders. 1999. Remarks on Holmberg’s Generalization. Studia Lin-guistica 53(1), 1–39.

Jackendoff, Ray S. 1975. Morphological and Semantic Regularities in theLexikon. Language 51(3), 639–671.

Jackendoff, Ray S. 2008. Construction after Construction and Its TheoreticalChallenges. Language 84(1), 8–28.

Jackendoff, Ray S. 2011. What is the human language faculty? Two views. Lan-guage 87(3), 586–624.

Jackendoff, Ray S. and Pinker, Steven. 2005. The Nature of the Language Facultyand its Implications for Evolution of Language (Reply to Fitch, Hauser, andChomsky). Cognition 97(2), 211–225.

Jacobs, Joachim. 2008. Wozu Konstruktionen? Linguistische Berichte 213, 3–44.Johnson, Mark. 1986. A GPSG Account of VP Structure in German. Linguistics

24(5), 871–882.Joshi, Aravind K., Levy, Leon S. and Takahashi, Masako. 1975. Tree Adjunct

Grammar. Journal of Computer and System Science 10(2), 136–163.Kallmeyer, Laura and Osswald, Rainer. To appear. A Frame-Based Semantics of

the Dative Alternation in Lexicalized Tree Adjoining Grammars. In ChristopherPiñón (ed.), Empirical Issues in Syntax and Semantics, volume 9, Paris: CNRS.

Kathol, Andreas. 1994. Passives without Lexical Rules. In Nerbonne et al. (1994),pages 237–272.

Kathol, Andreas. 1995. Linearization-Based German Syntax. Ph. D.thesis, OhioState University.

Kathol, Andreas. 2001. Positional Effects in a Monostratal Grammar of German.Journal of Linguistics 37(1), 35–66.

Kathol, Andreas. 2003. Subjects in Unexpected Places and the Notion of “Predi-cate”. In Pieter A. M. Seuren and Gerard Kempen (eds.), Verb Clusters in Ger-man and Dutch, Current Issues in Linguistic Theory, No. 242, Amsterdam andPhiladelphia: John Benjamins Publishing Co.

Kay, Paul. 2005. Argument Structure Constructions and the Argument-AdjunctDistinction. In Mirjam Fried and Hans C. Boas (eds.), Grammatical Construc-tions: Back to the Roots, Constructional Approaches to Language, No. 4, pages71–98, Amsterdam: John Benjamins Publishing Co. http://www.icsi.berkeley.edu/~kay/ASCs.pdf, 07.11.2006.

Kay, Paul and Fillmore, Charles J. 1999. Grammatical Constructions and Lin-guistic Generalizations: the What’s X Doing Y? Construction. Language 75(1),1–33.

Keil, Martina. 1997. Wort für Wort: Repräsentation und Verarbeitung verbalerPhraseologismen(Phraseo-Lex). Sprache und Information, No. 35, Tübingen:


http://www.icsi.berkeley.edu/~kay/ASCs.pdf

http://www.icsi.berkeley.edu/~kay/ASCs.pdf


REFERENCES 77

Max Niemeyer Verlag.King, Paul. 1999. Towards Truth in Head-Driven Phrase Structure Grammar.

In Valia Kordoni (ed.), Tübingen Studies in Head-Driven Phrase Struc-ture Grammar, Arbeitsberichte des SFB 340, No. No. 132, pages 301–352,Tübingen: Universität Tübingen. http://www.sfs.uni-tuebingen.de/sfb/reports/berichte/132/132abs.html, 12.10.2006.

Kiparsky, Paul. 1987. Morphology and Grammatical Relations. Unpublished pa-per, Stanford University, Stanford.

Kiparsky, Paul. 1988. Agreement and Linking Theory. Unpublished paper, Stan-ford University, Stanford.

Koenig, Jean-Pierre. 1999. Lexical Relations. Stanford Monographs in Linguis-tics, Stanford, CA: CSLI Publications.

Kordoni, Valia. 2001. Linking Experiencer-Subject Psych Verb Constructionsin Modern Greek. In Daniel P. Flickinger and Andreas Kathol (eds.), Pro-ceedings of the HPSG-2000 Conference, University of California, Berkeley,pages 198–213, CSLI Publications. http://cslipublications.stanford.edu/HPSG/1/, 11.10.2003.

Kratzer, Angelika. 1996. Severing the External Argument from Its Verb. In JohanRooryck and Laurie Zaring (eds.), Phrase Structure and the Lexicon, pages109–137, Dordrecht/Boston/London: Kluwer Academic Publishers.

Krenn, Brigitte and Erbach, Gregor. 1994. Idioms and Support Verb Construc-tions. In Nerbonne et al. (1994), pages 365–396.

Kroch, Anthony S. and Joshi, Aravind K. 1985. The Linguistic Relevance of TreeAdjoining Grammar. Technical Report MS-CIS-85-16, University of Pennsyl-vania. ftp://babel.ling.upenn.edu/papers/faculty/tony_kroch/papers/relevance3.pdf, 21.03.2008.

Kunze, Jürgen. 1991. Kasusrelationen und semantische Emphase. studia gram-matica XXXII, Berlin: Akademie Verlag.

Kunze, Jürgen. 1993. Sememstrukturen und Feldstrukturen. studia grammaticaXXXVI, Berlin: Akademie Verlag, unter Mitarbeit von Beate Firzlaff.

Lakoff, George. 1969. On Generative Semantics. Reproduced by the Indiana Uni-versity Linguistics Club.

Langacker, Ronald W. 1987. Foundations of Cognitive Grammar, volume 1. Stan-ford, CA: Stanford University Press.

Levin, Beth and Rappaport Hovav, Malka. 2005. Argument Realization. Cam-bridge University Press.

Lewis, Geoffrey L. 1967. Turkish Grammar. Oxford: Clarendon Press.Lohndal, Terje. 2012. Toward the End of Argument Structure. In María Cristina

Cuervo and Yves Roberge (eds.), The End of Argument Structure?, volume 38of Syntax and Semantics, pages 155–184, Bingley, UK: Emerald Group Pub-lishing.


http://www.sfs.uni-tuebingen.de/sfb/reports/berichte/132/132abs.html

http://www.sfs.uni-tuebingen.de/sfb/reports/berichte/132/132abs.html



ftp://babel.ling.upenn.edu/papers/faculty/tony_kroch/papers/relevance3.pdf

ftp://babel.ling.upenn.edu/papers/faculty/tony_kroch/papers/relevance3.pdf


78 REFERENCES

Lüdeling, Anke. 2001. On Particle Verbs and Similar Constructions in German.Dissertations in Linguistics, Stanford, CA: CSLI Publications.

Lüdeling, Anke and Kytö, Merja (eds.). 2009. Corpus Linguistics. An Interna-tional Handbook, volume 29.2 of Handbücher zur Sprach- und Kommunika-tionswissenschaft. Berlin: Mouton de Gruyter.

Marantz, Alec. 1981. A Theory of Grammatical Relations. Ph. D.thesis, MIT.Marantz, Alec. 1984. On the Nature of Grammatical Relations. Linguistic Inquiry

Monographs, No. 10, Cambridge, MA: MIT Press.Marantz, Alec. 1997. No Escape from Syntax. Don’t Try Morphological Anal-

ysis in the Privacy of Your Own Lexicon. U. Penn Working Papers in Lin-guistics 4(2), 201–225. http://www.ling.upenn.edu/papers/v4.2-contents.html,30.05.2010.

Meurers, Walt Detmar. 2000. Lexical Generalizations in the Syntax ofGerman Non-Finite Constructions. Arbeitspapiere des SFB 340 No. 145,Eberhard-Karls-Universität, Tübingen. http://www.sfs.uni-tuebingen.de/~dm/papers/diss.html, 19.08.2002.

Meurers, Walt Detmar. 2001. On Expressing Lexical Generalizations in HPSG.Nordic Journal of Linguistics 24(2), 161–217. http://www.sfs.uni-tuebingen.de/~dm/papers/lexical-generalizations.html, 30.11.2006.

Meurers, Walt Detmar and De Kuthy, Kordula. 2001. Case Assignment in Par-tially Fronted Constituents. In Christian Rohrer, Antje Roßdeutscher and HansKamp (eds.), Linguistic Form and its Computation, CSLI Studies in Computa-tional Linguistics, No. 1, pages 29–63, Stanford, CA: CSLI Publications.

Meurers, Walt Detmar and Müller, Stefan. 2009. Corpora and Syntax. In Lüdelingand Kytö (2009), Chapter 42, pages 920–933.

Michaelis, Laura A. and Ruppenhofer, Josef. 2001. Beyond Alternations: A Con-structional Model of the German Applicative Pattern. Stanford Monographs inLinguistics, Stanford, CA: CSLI Publications.

Montague, Richard. 1973. The Proper Treatment of Quantification in OrdinaryEnglish. In R. H. Thomason (ed.), Formal Philosophy. Selected Papers ofRichard Montague, pages 247–270, Yale University Press.

Müller, Gereon. 1998. Incomplete Category Fronting. A Derivational Approachto Remnant Movement inGerman. Studies in Natural Language and LinguisticTheory, No. 42, Dordrecht/Boston/London: Kluwer Academic Publishers.

Müller, Gereon. 2011. Regeln oder Konstruktionen? Von verblosen Direktivenzur sequentiellen Nominalreduplikation. In Stefan Engelberg, Anke Hollerand Kristel Proost (eds.), Sprachliches Wissen zwischen Lexikon und Gram-matik, Institut für Deutsche Sprache, Jahrbuch 2010, pages 211–249, Berlin/New York, NY: de Gruyter. http://www.uni-leipzig.de/~muellerg/mu242.pdf,13.05.2010.

Müller, Stefan. 1996. Yet another Paper about Partial Verb Phrase Fronting in


http://www.ling.upenn.edu/papers/v4.2-contents.html

http://www.sfs.uni-tuebingen.de/~dm/papers/diss.html

http://www.sfs.uni-tuebingen.de/~dm/papers/diss.html

http://www.sfs.uni-tuebingen.de/~dm/papers/lexical-generalizations.html

http://www.sfs.uni-tuebingen.de/~dm/papers/lexical-generalizations.html

http://www.uni-leipzig.de/~muellerg/mu242.pdf


REFERENCES 79

German. In Tsuji (1996), pages 800–805. http://hpsg.fu-berlin.de/~stefan/Pub/pvp.html, 19.10.2012.

Müller, Stefan. 1999. Deutsche Syntax deklarativ. Head-Driven Phrase StructureGrammar für das Deutsche. Linguistische Arbeiten, No. 394, Tübingen: MaxNiemeyer Verlag. http://hpsg.fu-berlin.de/~stefan/Pub/hpsg.html, 19.10.2012.

Müller, Stefan. 2000. Complex Predicates: Verbal Complexes, ResultativeConstructions, and Particle Verbsin German. Habilitationsschrift, Universi-tät des Saarlandes, Saarbrücken. http://hpsg.fu-berlin.de/~stefan/Pub/complex.html, 19.10.2012.

Müller, Stefan. 2001. Case in German – Towards an HPSG Analysis. In Walt Det-mar Meurers and Tibor Kiss (eds.), Constraint-Based Approaches to GermanicSyntax, Studies in Constraint-Based Lexicalism, No. 7, pages 217–255, Stan-ford, CA: CSLI Publications. http://hpsg.fu-berlin.de/~stefan/Pub/case.html,19.10.2012.

Müller, Stefan. 2002a. Complex Predicates: Verbal Complexes, Resultative Con-structions, and Particle Verbsin German. Studies in Constraint-Based Lexical-ism, No. 13, Stanford, CA: CSLI Publications. http://hpsg.fu-berlin.de/~stefan/Pub/complex.html, 19.10.2012.

Müller, Stefan. 2002b. Syntax or Morphology: German Particle Verbs Revis-ited. In Dehé et al. (2002), pages 119–139. http://hpsg.fu-berlin.de/~stefan/Pub/syn-morph-part.html, 19.10.2012.

Müller, Stefan. 2003a. Object-To-Subject-Raising and Lexical Rule. An Analysisof the GermanPassive. In Stefan Müller (ed.), Proceedings of the InternationalConference on Head-Driven Phrase Structure Grammar, volume 4, pages 278–297, Stanford, CA: CSLI Publications. http://hpsg.fu-berlin.de/~stefan/Pub/passiv-da.html, 19.10.2012.

Müller, Stefan. 2003b. Solving the Bracketing Paradox: an Analysis of the Mor-phology of German Particle Verbs. Journal of Linguistics 39(2), 275–325.http://hpsg.fu-berlin.de/~stefan/Pub/paradox.html, 19.10.2012.

Müller, Stefan. 2005a. Zur Analyse der deutschen Satzstruktur. Linguistische Be-richte 201, 3–39. http://hpsg.fu-berlin.de/~stefan/Pub/satz-lb.html, 19.10.2012.

Müller, Stefan. 2005b. Zur Analyse der scheinbar mehrfachen Vorfeldbeset-zung. Linguistische Berichte 203, 297–330. http://hpsg.fu-berlin.de/~stefan/Pub/mehr-vf-lb.html, 19.10.2012.

Müller, Stefan. 2006. Phrasal or Lexical Constructions? Language 82(4), 850–883. http://hpsg.fu-berlin.de/~stefan/Pub/phrasal.html, 19.10.2012.

Müller, Stefan. 2007a. Phrasal or Lexical Constructions: Some Comments on Un-derspecification of Constituent Order, Compositionality, and Control. In Müller(2007b), pages 373–393. http://hpsg.fu-berlin.de/~stefan/Pub/phrasal-2.html,29.07.2007.

Müller, Stefan (ed.). 2007b. Proceedings of the International Conference on


http://hpsg.fu-berlin.de/~stefan/Pub/pvp.html

http://hpsg.fu-berlin.de/~stefan/Pub/pvp.html

http://hpsg.fu-berlin.de/~stefan/Pub/hpsg.html

http://hpsg.fu-berlin.de/~stefan/Pub/complex.html


http://hpsg.fu-berlin.de/~stefan/Pub/case.html



http://hpsg.fu-berlin.de/~stefan/Pub/syn-morph-part.html

http://hpsg.fu-berlin.de/~stefan/Pub/syn-morph-part.html

http://hpsg.fu-berlin.de/~stefan/Pub/passiv-da.html

http://hpsg.fu-berlin.de/~stefan/Pub/passiv-da.html

http://hpsg.fu-berlin.de/~stefan/Pub/paradox.html

http://hpsg.fu-berlin.de/~stefan/Pub/satz-lb.html

http://hpsg.fu-berlin.de/~stefan/Pub/mehr-vf-lb.html

http://hpsg.fu-berlin.de/~stefan/Pub/mehr-vf-lb.html

http://hpsg.fu-berlin.de/~stefan/Pub/phrasal.html

http://hpsg.fu-berlin.de/~stefan/Pub/phrasal-2.html


80 REFERENCES

Head-Driven Phrase Structure Grammar, Stanford, CA, CSLI Publications.http://cslipublications.stanford.edu/HPSG/8/, 29.07.2007.

Müller, Stefan. 2007c. Qualitative Korpusanalyse für die Grammatiktheorie:Introspektion vs. Korpus. In Gisela Zifonun and Werner Kallmeyer (eds.),Sprachkorpora – Datenmengen und Erkenntnisfortschritt, Institut für DeutscheSprache Jahrbuch 2006, pages 70–90, Berlin: Walter de Gruyter. http://hpsg.fu-berlin.de/~stefan/Pub/intro-korpus.html, 19.10.2012.

Müller, Stefan. 2008. Head-Driven Phrase Structure Grammar: Eine Einführung.Stauffenburg Einführungen, No. 17, Tübingen: Stauffenburg Verlag, secondedition. http://hpsg.fu-berlin.de/~stefan/Pub/hpsg-lehrbuch.html, 19.10.2012.

Müller, Stefan. 2010a. Grammatiktheorie. Stauffenburg Einführungen,No. 20, Tübingen: Stauffenburg Verlag. http://hpsg.fu-berlin.de/~stefan/Pub/grammatiktheorie.html, 19.10.2012.

Müller, Stefan. 2010b. Persian Complex Predicates and the Limits of Inheritance-Based Analyses. Journal of Linguistics 46(3), 601–655. http://hpsg.fu-berlin.de/~stefan/Pub/persian-cp.html, 19.10.2012.

Müller, Stefan and Ørsnes, Bjarne. 2012. Towards an HPSG Analysis of ObjectShift in Danish. Ms, Freie Universität Berlin. http://hpsg.fu-berlin.de/~stefan/Pub/object-shift.html, 19.10.2012.

Müller, Stefan and Ørsnes, Bjarne. To appear. Danish in Head-Driven PhraseStructure Grammar. Ms, Freie Universität Berlin. http://hpsg.fu-berlin.de/~stefan/Pub/danish.html, 19.10.2012.

Nerbonne, John. 1986. ‘Phantoms’ and German Fronting: Poltergeist Con-stituents? Linguistics 24(5), 857–870.

Nerbonne, John. 1994. Partial Verb Phrases and Spurious Ambiguities. In Ner-bonne et al. (1994), pages 109–150.

Nerbonne, John, Netter, Klaus and Pollard, Carl J. (eds.). 1994. German in Head-Driven Phrase Structure Grammar. CSLI Lecture Notes, No. 46, Stanford, CA:CSLI Publications.

Noonan, Michael. 1994. A Tale of Two Passives in Irish. In Barbara Fox andPaul J. Hopper (eds.), Voice: Form and Function, Typological Studies in Lan-guage, No. 27, page 279–311, Amsterdam and Philadelphia: John BenjaminsPublishing Co.

Nunberg, Geoffrey. 1995. Transfers of Meaning. Journal of semantics 12(2), 109–132.

Nunberg, Geoffrey, Sag, Ivan A. and Wasow, Thomas. 1994. Idioms. Language70(3), 491–538.

Ørsnes, Bjarne. 2009. Das Verbalfeldmodell – ein Stellungsfeldermodell für denkontrastiven DaF-Unterricht. Deutsch als Fremdsprache 46(3), 143–149.

Özkaragöz, Inci. 1986. Monoclausal Double Passives in Turkish. In Dan I. Slobinand Karl Zimmer (eds.), Studies in Turkish Linguistics, Typological Studies in



http://hpsg.fu-berlin.de/~stefan/Pub/intro-korpus.html

http://hpsg.fu-berlin.de/~stefan/Pub/intro-korpus.html

http://hpsg.fu-berlin.de/~stefan/Pub/hpsg-lehrbuch.html

http://hpsg.fu-berlin.de/~stefan/Pub/grammatiktheorie.html

http://hpsg.fu-berlin.de/~stefan/Pub/grammatiktheorie.html

http://hpsg.fu-berlin.de/~stefan/Pub/persian-cp.html

http://hpsg.fu-berlin.de/~stefan/Pub/persian-cp.html

http://hpsg.fu-berlin.de/~stefan/Pub/object-shift.html

http://hpsg.fu-berlin.de/~stefan/Pub/object-shift.html

http://hpsg.fu-berlin.de/~stefan/Pub/danish.html

http://hpsg.fu-berlin.de/~stefan/Pub/danish.html


REFERENCES 81

Language, No. 8, John Benjamins Publishing Co.Peeters, Bert. 2000. The lexicon–encyclopedia interface. Amsterdam: Elsevier.Perlmutter, David M. (ed.). 1983. Studies in Relational Grammar, volume 1. The

University of Chicago Press.Pesetsky, David. 1996. Zero syntax: Experiencers and cascades. Cambridge,

Massachusetts: MIT press.Pollard, Carl J. 1994. Toward a Unified Account of Passive in German. In Ner-

bonne et al. (1994), pages 273–296.Pollard, Carl J. 1996. On Head Non-Movement. In Harry Bunt and Arthur

van Horck (eds.), Discontinuous Constituency, Natural Language Processing,No. 6, pages 279–305, Berlin/New York, NY: Mouton de Gruyter, veröf-fentlichte Version eines Ms. von 1990.

Pollard, Carl J. and Sag, Ivan A. 1987. Information-Based Syntax and Semantics.CSLI Lecture Notes, No. 13, Stanford, CA: CSLI Publications.

Pollard, Carl J. and Sag, Ivan A. 1994. Head-Driven Phrase Structure Grammar.Studies in Contemporary Linguistics, Chicago, IL/London: The University ofChicago Press.

Pulvermüller, Friedemann, Cappelle, Bert and Shtyrov, Yury. To appear. Brain Ba-sis of Meaning, Words, Constructions, and Grammar. In Graeme Trousdale andThomas Hoffmann (eds.), The Oxford Handbook of Construction Grammar,Oxford: Oxford University Press.

Rappaport, Malka. 1983. On the Nature of Derived Nominals. In Papers in lexical-functional grammar, pages 113–42, Bloomington: Indiana University Linguis-tics Club.

Reape, Mike. 1994. Domain Union and Word Order Variation in German. In Ner-bonne et al. (1994), pages 151–198.

Reis, Marga. 1982. Zum Subjektbegriff im Deutschen. In Werner Abraham (ed.),Satzglieder im Deutschen – Vorschläge zur syntaktischen, semantischen und-pragmatischen Fundierung, Studien zur deutschen Grammatik, No. 15, pages171–211, Tübingen: originally Gunter Narr Verlag now Stauffenburg Verlag.

Richter, Frank and Sailer, Manfred. 2009. Phraseological Clauses as Construc-tions in HPSG. In Stefan Müller (ed.), Proceedings of the International Con-ference on Head-Driven Phrase Structure Grammar, pages 297–317, Stanford,CA: CSLI Publications.

Riehemann, Susanne. 1993. Word Formation in Lexical Type Hierarchies: A CaseStudy of bar-Adjectives in German. Masters Thesis, Eberhard-Karls-Universi-tät Tübingen, also published as SfS-Report-02-93, Seminar für Sprachwissen-schaft, University of Tübingen.

Riehemann, Susanne Z. 1998. Type-Based Derivational Morphology. Journal ofComparative Germanic Linguistics 2(1), 49–77. http://doors.stanford.edu/~sr/morphology.ps, 05.04.2009.


http://doors.stanford.edu/~sr/morphology.ps

http://doors.stanford.edu/~sr/morphology.ps


82 REFERENCES

Ross, John R. 1974. There, There, (There, (There, (There . . . ))). In Chicago Lin-guistic Society, volume 10, pages 569–587.

Sag, Ivan A. 2007. Remarks on Locality. In Müller (2007b), pages 394–414. http://cslipublications.stanford.edu/HPSG/8/, 29.07.2007.

Sag, Ivan A. To appear. Sign-Based Construction Grammar: An Informal Synop-sis. In Hans C. Boas and Ivan A. Sag (eds.), Sign-based Construction Gram-mar, Stanford, CA: CSLI Publications. http://lingo.stanford.edu/sag/papers/theo-syno.pdf, 23.10.2007.

Sailer, Manfred. 2000. Combinatorial Semantics and Idiomatic Expressionsin Head-Driven Phrase StructureGrammar. Dissertation, Eberhard-Karls-Universität Tübingen. http://w210.ub.uni-tuebingen.de/dbt/volltexte/2003/916/, 18.11.2004.

Schein, Barry. 1993. Plurals and Events. Current Studies in Linguistics, No. 23,Cambridge, MA: MIT Press.

Scherpenisse, Wim. 1986. The Connection Between Base Structure and Lineariza-tion Restrictions in Germanand Dutch. Europäische Hochschulschriften, ReiheXXI, Linguistik, No. 47, Frankfurt/M.: Peter Lang.

Shtyrov, Y., Pihko, E. and Pulvermüller, F. 2005. Determinants of Dominance: IsLanguage Laterality Explained by Physical or Linguistic Features of Speech?Neuroimage 27(1), 37–47.

Simpson, Jane. 1983. Resultatives. In Lori S. Levin, Malka Rappaport and An-nie Zaenen (eds.), Papers in Lexical Functional Grammar, Indiana UniversityLinguistics Club, Reprint: Simpson, 2005.

Simpson, Jane. 2005. Resultatives. In Miriam Butt and Tracy Holloway King(eds.), Lexical Semantics in LFG, pages 149–161, Stanford, CA: CSLI Pub-lications.

Smith, Carlota S. 1970. Jespersen’s “move and change” class and causative verbsin English. In Peeters (2000), pages 101–09.

Smith, Carlota S. 1972. On causative verbs and derived nominals in English. Lin-guistic Inquiry 3(1), 136–138.

Soehn, Jan-Philipp. 2006. Über Bärendienste und erstaunte Bauklötze – Idiomeohne freie Lesart in der HPSG. Deutsche Sprache und Literatur, No. 1930,Frankfurt/Main: Peter Lang.

Soehn, Jan-Philipp and Sailer, Manfred. 2008. At First Blush on Tenterhooks.About Selectional Restrictions Imposed by Nonheads. In Gerhard Jäger, PaolaMonachesi, Gerald Penn and Shuly Wintner (eds.), Proceedings of FormalGrammar 2003, Vienna, Austria, pages 149–161, Stanford, CA: CSLI Publi-cations. http://cslipublications.stanford.edu/FG/2003/soehn.pdf, 04.06.2010.

Steedman, Mark J. 1991. Structure and Intonation. Language 67(2), 260–296.Steedman, Mark J. 2000. The Syntactic Process. Language, Speech, and Commu-

nication, Cambridge, MA/London, England: MIT Press.




http://lingo.stanford.edu/sag/papers/theo-syno.pdf

http://lingo.stanford.edu/sag/papers/theo-syno.pdf

http://w210.ub.uni-tuebingen.de/dbt/volltexte/2003/916/

http://w210.ub.uni-tuebingen.de/dbt/volltexte/2003/916/

http://cslipublications.stanford.edu/FG/2003/soehn.pdf


REFERENCES 83

Steedman, Mark J. and Baldridge, Jason. 2006. Combinatory Categorial Gram-mar. In Keith Brown (ed.), Encyclopedia of Language and Linguistics, pages610–621, Oxford: Elsevier, second edition.

Stefanowitsch, Anatol and Gries, Stephan Th. 2009. Corpora and Grammar. InLüdeling and Kytö (2009), Chapter 43, pages 933–952.

Sternefeld, Wolfgang. 1985. Deutsch ohne grammatische Funktionen: Ein Beitragzur Rektions- undBindungstheorie. Linguistische Berichte 99, 394–439.

Stiebels, Barbara. 1996. Lexikalische Argumente und Adjunkte: Zum semantis-chen Beitrag verbaler Präfixeund Partikeln. studia grammatica XXXIX, Berlin:Akademie Verlag.

Timberlake, Alan. 1982. The Impersonal Passive in Lithuanian. In Proceedings ofthe Eighth Annual Meeting of the Berkeley Linguistics Society, pages 508–524,Berkeley: University of California.

Tomasello, Michael. 2003. Constructing a Language: A Usage-Based Theory ofLanguage Acquisition. Cambridge, MA: Harvard University Press.

Tomasello, Michael, Akhtar, Nameera, Dodsen, Kelly and Rekau, Laura. 1997.Differential Productivity in Young Children’s Use of Nouns and Verbs. Journalof Child Language 24(2), 373–387.

Tsuji, Jun-ichi (ed.). 1996. Proceedings of Coling-96. 16th International Confer-ence on ComputationalLinguistics (COLING96). Copenhagen, Denmark, Au-gust 5–9, 1996, Copenhagen, Denmark, Association for Computational Lin-guistics.

Uszkoreit, Hans. 1987. Word Order and Constituent Structure in German. CSLILecture Notes, No. 8, Stanford, CA: CSLI Publications.

Verspoor, Cornelia Maria. 1997. Contextually-Dependent Lexical Semantics.Ph. D.thesis, University of Edinburgh. ftp://ftp.cogsci.ed.ac.uk/pub/kversp/thesis.ps.gz, 04.06.2010.

Webelhuth, Gert. 1995. X-bar Theory and Case Theory. In Gert Webelhuth (ed.),Government and Binding Theory and the Minimalist Program. Principles andParametersin Syntactic Theory, Generative Syntax, pages 15–95, Oxford, UK& Cambrigde, USA: Blackwell Publishing Ltd.

Wechsler, Stephen. 2008a. Dualist Syntax. In Stefan Müller (ed.), Proceed-ings of the International Conference on Head-Driven Phrase Structure Gram-mar, pages 294–304, Stanford, CA: CSLI Publications. http://cslipublications.stanford.edu/HPSG/9/, 31.10.2008.

Wechsler, Stephen Mark. 1991. Argument Structure and Linking. Ph. D.thesis,Stanford University.

Wechsler, Stephen Mark. 1995. The Semantic Basis of Argument Structure. Dis-sertations in Linguistics, Stanford: CSLI Publications.

Wechsler, Stephen Mark. 1997. Resultative Predicates and Control. In Ralph C.Blight and Michelle J. Moosally (eds.), Texas Linguistic Forum 38: The Syntax


ftp://ftp.cogsci.ed.ac.uk/pub/kversp/thesis.ps.gz

ftp://ftp.cogsci.ed.ac.uk/pub/kversp/thesis.ps.gz




84 REFERENCES

and Semantics of Predication. Proceedings ofthe 1997 Texas Linguistics SocietyConference, pages 307–321, Austin, Texas: University of Texas Department ofLinguistics.

Wechsler, Stephen Mark. 2005. What is Right and Wrong about Little v. In Gram-mar and Beyond—Essays in Honour of Lars Hellan, pages 179–195, Oslo, Nor-way: Novus Press.

Wechsler, Stephen Mark. 2008b. A Diachronic Account of English DeverbalNominals. In Charles B. Chang and Hannah J. Haynie (eds.), Proceedings of the26th West Coast Conference on Formal Linguistics, pages 498–506, Somerville,MA: Cascadilla Proceedings Project.

Wechsler, Stephen Mark and Noh, Bokyung. 2001. On Resultative Predicates andClauses: Parallels between Korean and English. Language Sciences 23(4), 391–423.

Wells, Rulon S. 1947. Immediate Constituents. Language 23(2), 81–117.Wetta, Andrew C. 2011. A Construction-based Cross-linguistic Analysis of V2

Word Order. In Stefan Müller (ed.), Proceedings of the 18th International Con-ference on Head-Driven Phrase Structure Grammar, University of Washington,pages 248–268, Stanford, CA: CSLI Publications.

Zaenen, Annie, Maling, Joan and Thráinsson, Höskuldur. 1985. Case andGrammatical Functions: The Icelandic Passive. Natural Language and Lin-guistic Theory 3(4), 441–483. http://www2.parc.com/istl/members/zaenen/publications/Passive.pdf, 04.12.2006.

Zucchi, Alessandro. 1993. The Language of Propositions and Events: Issues in theSyntax and the Semantics of Nominalization. Studies in Linguistics and Philos-ophy, No. 51, Berlin/Heidelberg/New York, NY: Springer Verlag.


http://www2.parc.com/istl/members/zaenen/publications/Passive.pdf

http://www2.parc.com/istl/members/zaenen/publications/Passive.pdf


Date post:	10-Apr-2018
Category:	Documents
Upload:	ngodang
View:	215 times
Download:	3 times

Lexical Approaches to Argument Structure - hu-berlin.destefan/PS/arg-st.pdf · Lexical Approaches...

Documents